[{"data":1,"prerenderedAt":6461},["ShallowReactive",2],{"blog-posts":3},[4,518,1643,3261,4095,4658,5187,5607,5827,6106],{"id":5,"title":6,"author":7,"body":8,"date":504,"description":505,"draft":506,"extension":507,"image":508,"meta":509,"navigation":510,"path":511,"seo":512,"stem":513,"tags":514,"__hash__":517},"blog/blog/nvidia-local-ai-openclaw-agent-platform.md","NVIDIA, Local AI, and Why OpenClaw Is Starting to Matter a Lot More","ClawNest Team",{"type":9,"value":10,"toc":488},"minimark",[11,15,18,21,24,40,43,48,51,54,57,60,63,70,74,77,80,83,89,92,95,98,101,105,108,111,133,136,139,142,146,149,152,158,161,164,172,175,178,182,185,191,194,211,214,217,220,223,227,239,242,247,250,253,256,262,266,269,272,275,278,281,287,290,297,301,304,307,310,333,336,339,342,345,349,352,355,358,361,364,367,373,376,380,383,386,389,392,395,401,405,411,414,417,443,449,455,459,462,465,477,479,482,485],[12,13,14],"p",{},"Something important is happening in AI right now.",[12,16,17],{},"The conversation is shifting away from pure model hype and toward systems that actually work. Not just models that score well in benchmarks. Real systems. Systems that can use tools, manage context, run across devices, connect to browsers, interact with software, and operate safely enough to be trusted with meaningful tasks.",[12,19,20],{},"That shift is exactly why OpenClaw matters.",[12,22,23],{},"And it is also why the latest OpenClaw releases feel more significant than they might look at first glance.",[12,25,26,27,31,32,35,36,39],{},"Versions 2026.3.22, 2026.3.23, and 2026.3.24 landed close together, but they are best understood as one combined move. Together, they push OpenClaw further into an important position in the stack: not as just another AI app, but as a serious ",[28,29,30],"strong",{},"agent platform"," for the world that is now forming around ",[28,33,34],{},"NVIDIA",", ",[28,37,38],{},"local AI",", and hybrid inference.",[12,41,42],{},"At ClawNest, this is the direction we care about most.",[44,45,47],"h2",{"id":46},"the-real-ai-opportunity-is-no-longer-just-the-model","The Real AI Opportunity Is No Longer Just the Model",[12,49,50],{},"For a while, the industry acted as if the whole game was about the smartest model.",[12,52,53],{},"That was always incomplete.",[12,55,56],{},"The real opportunity is building systems around models. The runtime. The orchestration layer. The permissions. The interfaces. The memory. The tool access. The device integration. The recovery paths. The security boundaries.",[12,58,59],{},"That is the messy layer between \"the model can do this in a demo\" and \"this product is useful every day.\"",[12,61,62],{},"It is also where most of the value will be built.",[12,64,65,66,69],{},"This is where ",[28,67,68],{},"OpenClaw"," is getting dramatically stronger.",[44,71,73],{"id":72},"why-nvidia-matters-so-much","Why NVIDIA Matters So Much",[12,75,76],{},"When people say NVIDIA, they usually mean GPUs.",[12,78,79],{},"That is too small a view.",[12,81,82],{},"NVIDIA now represents the wider acceleration of practical AI infrastructure. Faster local inference. Better edge deployment options. More serious workstation AI. More ambitious on prem setups. More experiments where users want control over data, latency, and cost.",[12,84,85,86,88],{},"As ",[28,87,38],{}," gets better, the question becomes obvious:",[12,90,91],{},"What is the operating layer that turns local models into useful systems?",[12,93,94],{},"Not just a chat box. Not just an API endpoint. A real runtime for agents.",[12,96,97],{},"That is the role OpenClaw is growing into.",[12,99,100],{},"If NVIDIA is helping define the compute layer for the next phase of AI, then platforms like OpenClaw are defining how that compute becomes usable.",[44,102,104],{"id":103},"these-releases-were-not-random","These Releases Were Not Random",[12,106,107],{},"Looking across 2026.3.22 through 2026.3.24, there is a very clear pattern.",[12,109,110],{},"OpenClaw is becoming:",[112,113,114,118,121,124,127,130],"ul",{},[115,116,117],"li",{},"more secure by default",[115,119,120],{},"more modular and plugin native",[115,122,123],{},"more compatible with OpenAI style ecosystems",[115,125,126],{},"better at hybrid local and cloud workflows",[115,128,129],{},"more polished as a product, not just a framework",[115,131,132],{},"more credible as the control plane for agents",[12,134,135],{},"That combination matters.",[12,137,138],{},"Because the future is not one giant monolith. It is not one model vendor. It is not one interface. It is a layered AI stack, and the winning products will sit in the layer that coordinates everything else.",[12,140,141],{},"That is the layer OpenClaw is claiming.",[44,143,145],{"id":144},"security-first-is-the-only-serious-position","Security First Is the Only Serious Position",[12,147,148],{},"One of the strongest signals in these releases is just how much hardening work went in.",[12,150,151],{},"There are improvements across exec approvals, webhook verification, proxy handling, media path restrictions, auth scopes, device pairing, browser access, sandbox environments, and more.",[12,153,154,155,157],{},"This is not glamorous work, but it is exactly the work that separates a toy from an ",[28,156,30],{},".",[12,159,160],{},"I think a lot of the AI industry still underestimates this. People get excited when an agent can click buttons or run commands. They spend less time thinking about what happens when that agent has ambiguous permissions, weak approval boundaries, or sloppy network controls.",[12,162,163],{},"That is backwards.",[12,165,85,166,168,169,171],{},[28,167,38],{}," gets faster on ",[28,170,34],{}," hardware, mistakes become more dangerous, not less. Better inference speed without better control is not progress. It is just a faster path to bad outcomes.",[12,173,174],{},"The recent OpenClaw releases show the right instinct: capability should grow together with restraint.",[12,176,177],{},"That is one of the reasons ClawNest is built around OpenClaw in the first place.",[44,179,181],{"id":180},"openclaw-is-becoming-a-real-agent-platform","OpenClaw Is Becoming a Real Agent Platform",[12,183,184],{},"This is the bigger story.",[12,186,187,188,190],{},"An ",[28,189,30],{}," should not be judged only by how smart the base model is. It should be judged by whether it can actually host durable, extensible, multi surface AI systems.",[12,192,193],{},"That means:",[112,195,196,199,202,205,208],{},[115,197,198],{},"plugins that can be maintained cleanly",[115,200,201],{},"tools that can be discovered and governed properly",[115,203,204],{},"browser and device access that does not feel bolted on",[115,206,207],{},"support for local, remote, and hybrid runtimes",[115,209,210],{},"compatibility with the APIs and conventions the rest of the ecosystem already uses",[12,212,213],{},"The recent OpenClaw releases made meaningful progress on all of that.",[12,215,216],{},"There were major improvements in sandboxing, plugin SDK boundaries, ClawHub flows, browser support, runtime compatibility, Control UI clarity, skills installation, and tool visibility.",[12,218,219],{},"None of these changes alone is the full story.",[12,221,222],{},"Together, they are.",[44,224,226],{"id":225},"better-openai-compatibility-is-a-big-strategic-move","Better OpenAI Compatibility Is a Big Strategic Move",[12,228,229,230,234,235,238],{},"One of the smartest recent changes is broader OpenAI compatible gateway support, including ",[231,232,233],"code",{},"/v1/models"," and ",[231,236,237],{},"/v1/embeddings",", plus better handling of explicit model overrides across chat and responses flows.",[12,240,241],{},"This matters because a huge part of the AI ecosystem now assumes OpenAI shaped interfaces, even when the actual backend is something else entirely.",[12,243,244,245,157],{},"That includes local stacks, retrieval systems, custom apps, internal tools, and emerging workflows built around ",[28,246,38],{},[12,248,249],{},"In practice, this means OpenClaw becomes easier to place in front of many different providers and runtimes without forcing every client to understand the differences underneath.",[12,251,252],{},"That is not flashy, but it is incredibly important.",[12,254,255],{},"Interoperability is a force multiplier.",[12,257,258,259,261],{},"And in a world where ",[28,260,34],{}," powered systems increasingly mix local inference with cloud services, interoperability is one of the most valuable things you can ship.",[44,263,265],{"id":264},"local-ai-needs-orchestration-not-just-inference","Local AI Needs Orchestration, Not Just Inference",[12,267,268],{},"This is the part I think the market still misses.",[12,270,271],{},"A lot of people talk about local AI as if the hard part is simply getting the model to run on your machine.",[12,273,274],{},"That is only step one.",[12,276,277],{},"The real challenge is everything after that.",[12,279,280],{},"How does the system access tools?\nHow does it browse?\nHow does it message?\nHow does it handle memory?\nHow does it recover from partial failures?\nHow does it work across mobile, desktop, browser, and chat surfaces?\nHow do you keep it useful without making it reckless?",[12,282,283,284,286],{},"That is why ",[28,285,68],{}," is interesting.",[12,288,289],{},"It is not trying to compete with NVIDIA. It is not trying to replace model providers. It is solving a different problem: making all of those components behave like one coherent product.",[12,291,292,293,296],{},"At ",[28,294,295],{},"ClawNest",", that is the opportunity we see most clearly.",[44,298,300],{"id":299},"clawnest-and-the-practical-ai-stack","ClawNest and the Practical AI Stack",[12,302,303],{},"ClawNest exists because we think the future of AI will belong to platforms that make complex systems feel usable.",[12,305,306],{},"Not just powerful. Usable.",[12,308,309],{},"That means taking the best parts of the modern stack:",[112,311,312,315,318,321,324,327,330],{},[115,313,314],{},"fast models",[115,316,317],{},"local AI where it makes sense",[115,319,320],{},"cloud models where they are stronger",[115,322,323],{},"browser and device access",[115,325,326],{},"messaging and multi channel delivery",[115,328,329],{},"memory and workflows",[115,331,332],{},"policy and guardrails",[12,334,335],{},"And turning them into something people can actually rely on.",[12,337,338],{},"That is why OpenClaw's recent momentum matters to us.",[12,340,341],{},"The project is maturing in the right places. The architecture is getting cleaner. The UX is improving. The security posture is getting more serious. The plugin story is getting stronger. The surrounding ecosystem is becoming more practical.",[12,343,344],{},"For ClawNest, that translates directly into a better platform for people who want AI that is not trapped in a demo.",[44,346,348],{"id":347},"product-quality-is-becoming-a-competitive-advantage","Product Quality Is Becoming a Competitive Advantage",[12,350,351],{},"Another thing I like in these releases is the amount of product polish.",[12,353,354],{},"There are better Control UI flows, better skills management, better markdown previews, better tool visibility, better session navigation, better runtime clarity, and better install and setup ergonomics.",[12,356,357],{},"This stuff matters.",[12,359,360],{},"A lot of open source AI infrastructure still behaves like it resents the user. It assumes that if something is powerful enough, rough edges do not matter.",[12,362,363],{},"I disagree.",[12,365,366],{},"Good infrastructure should still feel good to use.",[12,368,369,370,372],{},"If an ",[28,371,30],{}," is going to sit between humans and powerful AI systems, the interface layer has to explain what is happening. It has to reduce ambiguity. It has to make configuration understandable. It has to help users trust the system.",[12,374,375],{},"OpenClaw is getting better at that.",[44,377,379],{"id":378},"the-multi-surface-future-is-already-here","The Multi Surface Future Is Already Here",[12,381,382],{},"One thing the recent releases make obvious is that OpenClaw is not being built for one interface.",[12,384,385],{},"There were updates across Microsoft Teams, Slack, Telegram, WhatsApp, Discord, Matrix, Android, browser tooling, Control UI, and desktop surfaces.",[12,387,388],{},"That is exactly right.",[12,390,391],{},"AI is not going to live in one app.",[12,393,394],{},"It is going to live across chat, browser, mobile, dashboards, background jobs, tools, and devices. The systems that win will be the ones that can move across those surfaces without becoming incoherent.",[12,396,397,398,400],{},"This is another reason ",[28,399,68],{}," stands out. It is increasingly designed as the connective layer across all of that.",[44,402,404],{"id":403},"why-this-matters-for-nvidia-and-local-ai-builders","Why This Matters for NVIDIA and Local AI Builders",[12,406,407,408,410],{},"If you are building around ",[28,409,34],{},", here is the practical takeaway.",[12,412,413],{},"The next wave of value is not just in better local model performance. It is in better systems around local model performance.",[12,415,416],{},"You need:",[112,418,419,422,425,428,431,434,437,440],{},[115,420,421],{},"orchestration",[115,423,424],{},"policy",[115,426,427],{},"tool routing",[115,429,430],{},"compatibility",[115,432,433],{},"memory",[115,435,436],{},"interfaces",[115,438,439],{},"safe execution",[115,441,442],{},"reliable delivery",[12,444,445,446,448],{},"That is where ",[28,447,68],{}," is becoming much more compelling.",[12,450,451,452,454],{},"And if you are using ",[28,453,295],{},", this is exactly the direction the platform is built for: taking the strength of modern AI infrastructure and turning it into something operational, coherent, and usable in the real world.",[44,456,458],{"id":457},"final-thought","Final Thought",[12,460,461],{},"The last three OpenClaw releases should not be read as a random stream of patch notes.",[12,463,464],{},"They should be read as a signal.",[12,466,467,468,470,471,473,474,476],{},"OpenClaw is growing into a more serious ",[28,469,30],{}," at exactly the moment when ",[28,472,34],{}," is helping push ",[28,475,38],{}," from experimentation toward practical deployment.",[12,478,135],{},[12,480,481],{},"Better compute needs better orchestration.\nBetter local models need better control planes.\nMore capable agents need better boundaries.",[12,483,484],{},"That is the direction OpenClaw is moving.",[12,486,487],{},"And at ClawNest, we think that is the right place to build.",{"title":489,"searchDepth":490,"depth":490,"links":491},"",2,[492,493,494,495,496,497,498,499,500,501,502,503],{"id":46,"depth":490,"text":47},{"id":72,"depth":490,"text":73},{"id":103,"depth":490,"text":104},{"id":144,"depth":490,"text":145},{"id":180,"depth":490,"text":181},{"id":225,"depth":490,"text":226},{"id":264,"depth":490,"text":265},{"id":299,"depth":490,"text":300},{"id":347,"depth":490,"text":348},{"id":378,"depth":490,"text":379},{"id":403,"depth":490,"text":404},{"id":457,"depth":490,"text":458},"2026-03-27","How OpenClaw is becoming a serious agent platform as NVIDIA accelerates local AI infrastructure. Covers the latest OpenClaw releases, security hardening, OpenAI compatibility, and what it means for the practical AI stack.",false,"md","/blog/clawnest_openclaw_nvidia.png",{},true,"/blog/nvidia-local-ai-openclaw-agent-platform",{"title":6,"description":505},"blog/nvidia-local-ai-openclaw-agent-platform",[68,34,515,516,295],"Local AI","Agent Platform","dO8GvUpQLyzTzxdOwQjWNqnn0Lwgs3njGIOvRtWO1Ss",{"id":519,"title":520,"author":7,"body":521,"date":1630,"description":1631,"draft":506,"extension":507,"image":1632,"meta":1633,"navigation":510,"path":1634,"seo":1635,"stem":1636,"tags":1637,"__hash__":1642},"blog/blog/nvidia-nemoclaw-openclaw-security.md","NVIDIA NemoClaw: Enterprise Security Comes to OpenClaw",{"type":9,"value":522,"toc":1612},[523,527,537,540,544,547,553,571,582,586,593,600,620,623,627,638,641,669,672,676,681,684,704,712,715,719,726,752,755,759,765,791,794,884,887,905,908,912,919,1137,1140,1144,1147,1232,1235,1239,1242,1420,1423,1427,1434,1437,1443,1446,1472,1475,1479,1482,1502,1505,1509,1512,1532,1535,1539,1546,1577,1580,1590,1593,1608],[44,524,526],{"id":525},"nemoclaw-nvidia-brings-enterprise-security-to-openclaw","NemoClaw: NVIDIA Brings Enterprise Security to OpenClaw",[12,528,529,530,533,534,157],{},"On March 16, 2026, NVIDIA CEO Jensen Huang announced ",[28,531,532],{},"NemoClaw"," at GTC 2026: an open-source security and privacy stack purpose-built for OpenClaw. It's the first major enterprise-focused extension to the OpenClaw ecosystem, and it directly addresses the biggest concern organizations have when deploying AI assistants: ",[28,535,536],{},"keeping sensitive data under control",[12,538,539],{},"NemoClaw is available today on GitHub under an Apache 2.0 license.",[44,541,543],{"id":542},"the-nvidia-ai-ecosystem","The NVIDIA AI Ecosystem",[12,545,546],{},"To understand NemoClaw's significance, it helps to see where it fits in NVIDIA's broader AI software stack.",[12,548,549,552],{},[28,550,551],{},"NeMo Framework"," is NVIDIA's end-to-end platform for training, fine-tuning, and deploying large language models. NemoClaw builds directly on two NeMo components:",[112,554,555,561],{},[115,556,557,560],{},[28,558,559],{},"NeMo Guardrails"," is the programmable safety layer that lets developers define conversational boundaries, topic restrictions, and output validation. NemoClaw extends Guardrails with OpenClaw-specific policies for tool execution and file access.",[115,562,563,566,567,570],{},[28,564,565],{},"TensorRT-LLM"," is NVIDIA's inference optimization engine that delivers ",[28,568,569],{},"2-4x throughput improvements"," over standard PyTorch serving. NemoClaw's local inference uses TensorRT-LLM under the hood, which is why Nemotron models run significantly faster on NVIDIA hardware than generic GGUF quantizations.",[12,572,573,574,577,578,581],{},"The ",[28,575,576],{},"NGC Catalog"," serves as the distribution channel for Nemotron models. NemoClaw pulls pre-optimized model configurations directly from NGC, including TensorRT-LLM engine files tuned for specific GPU architectures. This means ",[231,579,580],{},"nemoclaw init"," automatically selects the best model variant for your hardware, with no manual optimization required.",[44,583,585],{"id":584},"gtc-2026-the-bigger-picture","GTC 2026: The Bigger Picture",[12,587,588,589,592],{},"GTC 2026's overarching theme was ",[28,590,591],{},"\"AI in Production\"",": the shift from prototyping to deploying AI systems at scale with enterprise requirements for reliability, security, and governance.",[12,594,595,596,599],{},"NemoClaw was presented during the ",[28,597,598],{},"open-source segment"," of Jensen Huang's keynote, alongside several other announcements that shape the hardware and software landscape NemoClaw targets:",[112,601,602,608,614],{},[115,603,604,607],{},[28,605,606],{},"DGX Spark & DGX Station"," are desktop-class AI systems with Grace Blackwell hardware, making local inference with larger Nemotron models practical outside the data center.",[115,609,610,613],{},[28,611,612],{},"AI Enterprise 6.0"," is NVIDIA's commercial AI platform, which now includes lifecycle management, support SLAs, and compliance tooling for production deployments.",[115,615,616,619],{},[28,617,618],{},"Blackwell Ultra & Vera Rubin architecture preview"," showcases next-generation GPU and CPU architectures that will expand the performance envelope for local AI workloads.",[12,621,622],{},"NemoClaw's announcement in the open-source segment, rather than the enterprise segment, was a deliberate signal: NVIDIA is investing in the open AI assistant ecosystem, not just proprietary solutions.",[44,624,626],{"id":625},"what-is-nemoclaw","What Is NemoClaw?",[12,628,629,630,633,634,637],{},"NemoClaw is a collection of open-source tools that wraps around an existing OpenClaw installation to add enterprise-grade privacy routing, sandboxed execution, and local inference capabilities. It combines NVIDIA's ",[28,631,632],{},"Nemotron"," language models with a hardened runtime called ",[28,635,636],{},"OpenShell"," to create a security layer that sits between your OpenClaw gateway and the outside world.",[12,639,640],{},"The entire stack installs with a single command:",[642,643,647],"pre",{"className":644,"code":645,"language":646,"meta":489,"style":489},"language-bash shiki shiki-themes github-light github-dark","nemoclaw init --config enterprise.yaml\n","bash",[231,648,649],{"__ignoreMap":489},[650,651,654,658,662,666],"span",{"class":652,"line":653},"line",1,[650,655,657],{"class":656},"sScJk","nemoclaw",[650,659,661],{"class":660},"sZZnC"," init",[650,663,665],{"class":664},"sj4cs"," --config",[650,667,668],{"class":660}," enterprise.yaml\n",[12,670,671],{},"This sets up the Privacy Router, OpenShell Runtime, local Nemotron inference (if supported hardware is detected), and the declarative policy engine. Everything is configured from one YAML blueprint.",[44,673,675],{"id":674},"key-features","Key Features",[677,678,680],"h3",{"id":679},"privacy-router","Privacy Router",[12,682,683],{},"The Privacy Router is the centerpiece of NemoClaw. It acts as an intelligent proxy between your OpenClaw gateway and cloud LLM providers, making real-time decisions about where each request should be processed:",[112,685,686,692,698],{},[115,687,688,691],{},[28,689,690],{},"Hybrid routing",": sensitive queries are automatically routed to local Nemotron models, while routine requests go to your configured cloud provider (Anthropic, OpenAI).",[115,693,694,697],{},[28,695,696],{},"PII stripping",": personally identifiable information is detected and redacted before any data leaves your network.",[115,699,700,703],{},[28,701,702],{},"Classification engine",": each message is scored for sensitivity using a lightweight local classifier, with configurable thresholds.",[642,705,710],{"className":706,"code":708,"language":709},[707],"language-text","User Message\n    ↓\nPrivacy Router (sensitivity classification)\n    ↓                    ↓\nLocal Nemotron       Cloud LLM (PII stripped)\n    ↓                    ↓\nOpenClaw Gateway (merged response)\n","text",[231,711,708],{"__ignoreMap":489},[12,713,714],{},"Organizations can define their own sensitivity rules. For example, you can route all messages containing financial data, medical records, or internal project names to local inference only.",[677,716,718],{"id":717},"openshell-runtime","OpenShell Runtime",[12,720,721,722,725],{},"OpenShell replaces OpenClaw's default shell execution with a ",[28,723,724],{},"kernel-level sandboxed environment",". Every tool invocation, script execution, and file operation runs inside a hardened container with:",[112,727,728,734,740,746],{},[115,729,730,733],{},[28,731,732],{},"Namespace isolation",": each execution gets its own process, network, and filesystem namespace.",[115,735,736,739],{},[28,737,738],{},"Policy-based access control",": fine-grained rules define which tools can access which resources.",[115,741,742,745],{},[28,743,744],{},"Audit logging",": every action is logged with full context for compliance review.",[115,747,748,751],{},[28,749,750],{},"Network filtering",": outbound connections are restricted to explicitly allowed endpoints.",[12,753,754],{},"This means even if a skill or tool behaves unexpectedly, it cannot access files, networks, or processes outside its allowed scope.",[677,756,758],{"id":757},"local-inference-with-nemotron","Local Inference with Nemotron",[12,760,761,762,764],{},"NemoClaw ships with optimized configurations for NVIDIA's ",[28,763,632],{}," model family, enabling fully local inference for sensitive workloads:",[112,766,767,773,779,785],{},[115,768,769,772],{},[28,770,771],{},"Zero token costs"," for locally processed queries. No API calls, no usage charges.",[115,774,775,778],{},[28,776,777],{},"Air-gapped operation"," ensures sensitive data never leaves the local network.",[115,780,781,784],{},[28,782,783],{},"Quantized models"," in INT4/INT8 variants are optimized for desktop-class GPUs.",[115,786,787,790],{},[28,788,789],{},"Streaming support"," provides real-time token generation with the same UX as cloud models.",[12,792,793],{},"The Nemotron family includes four tiers, each targeting different hardware profiles:",[795,796,797,816],"table",{},[798,799,800],"thead",{},[801,802,803,807,810,813],"tr",{},[804,805,806],"th",{},"Model",[804,808,809],{},"Parameters",[804,811,812],{},"Target Hardware",[804,814,815],{},"Use Case",[817,818,819,836,852,868],"tbody",{},[801,820,821,827,830,833],{},[822,823,824],"td",{},[28,825,826],{},"Nemotron-Mini",[822,828,829],{},"8B",[822,831,832],{},"GeForce RTX 4090/5090",[822,834,835],{},"Classification, short Q&A",[801,837,838,843,846,849],{},[822,839,840],{},[28,841,842],{},"Nemotron-22B",[822,844,845],{},"22B",[822,847,848],{},"RTX PRO 6000",[822,850,851],{},"General assistant tasks, summarization",[801,853,854,859,862,865],{},[822,855,856],{},[28,857,858],{},"Nemotron-49B",[822,860,861],{},"49B",[822,863,864],{},"DGX Spark",[822,866,867],{},"Complex reasoning, code generation",[801,869,870,875,878,881],{},[822,871,872],{},[28,873,874],{},"Nemotron-70B",[822,876,877],{},"70B",[822,879,880],{},"DGX Station",[822,882,883],{},"Full-capability local inference",[12,885,886],{},"NemoClaw uses two quantization strategies to fit these models on different hardware:",[112,888,889,895],{},[115,890,891,894],{},[28,892,893],{},"AWQ (INT4)"," uses Activation-aware Weight Quantization for consumer GPUs. Nemotron-Mini INT4 fits in ~6 GB VRAM while retaining strong classification accuracy.",[115,896,897,900,901,904],{},[28,898,899],{},"SmoothQuant (INT8)"," balances quality and speed for professional GPUs. Nemotron-22B INT8 on an RTX PRO 6000 achieves approximately ",[28,902,903],{},"45 tokens/s",", fast enough for real-time streaming responses.",[12,906,907],{},"The local models handle classification, summarization, and general Q&A tasks. Complex reasoning tasks can still be routed to cloud providers (with PII stripping) when needed.",[677,909,911],{"id":910},"declarative-policy-engine","Declarative Policy Engine",[12,913,914,915,918],{},"All security configuration lives in ",[28,916,917],{},"versioned YAML blueprints"," that define the complete security posture of your deployment:",[642,920,924],{"className":921,"code":922,"language":923,"meta":489,"style":489},"language-yaml shiki shiki-themes github-light github-dark","# nemoclaw-policy.yaml\nprivacy:\n  pii_detection: strict\n  routing:\n    sensitive: local\n    default: cloud\n\nruntime:\n  sandbox: openshell\n  network:\n    allow:\n      - \"api.anthropic.com\"\n      - \"api.openai.com\"\n    deny: [\"*\"]\n  filesystem:\n    read: [\"/data/knowledge-base\"]\n    write: [\"/tmp/openclaw\"]\n\naudit:\n  level: full\n  retention: 90d\n  export: siem\n","yaml",[231,925,926,932,942,954,962,973,984,990,998,1009,1017,1025,1034,1042,1057,1065,1078,1091,1096,1104,1115,1126],{"__ignoreMap":489},[650,927,928],{"class":652,"line":653},[650,929,931],{"class":930},"sJ8bj","# nemoclaw-policy.yaml\n",[650,933,934,938],{"class":652,"line":490},[650,935,937],{"class":936},"s9eBZ","privacy",[650,939,941],{"class":940},"sVt8B",":\n",[650,943,945,948,951],{"class":652,"line":944},3,[650,946,947],{"class":936},"  pii_detection",[650,949,950],{"class":940},": ",[650,952,953],{"class":660},"strict\n",[650,955,957,960],{"class":652,"line":956},4,[650,958,959],{"class":936},"  routing",[650,961,941],{"class":940},[650,963,965,968,970],{"class":652,"line":964},5,[650,966,967],{"class":936},"    sensitive",[650,969,950],{"class":940},[650,971,972],{"class":660},"local\n",[650,974,976,979,981],{"class":652,"line":975},6,[650,977,978],{"class":936},"    default",[650,980,950],{"class":940},[650,982,983],{"class":660},"cloud\n",[650,985,987],{"class":652,"line":986},7,[650,988,989],{"emptyLinePlaceholder":510},"\n",[650,991,993,996],{"class":652,"line":992},8,[650,994,995],{"class":936},"runtime",[650,997,941],{"class":940},[650,999,1001,1004,1006],{"class":652,"line":1000},9,[650,1002,1003],{"class":936},"  sandbox",[650,1005,950],{"class":940},[650,1007,1008],{"class":660},"openshell\n",[650,1010,1012,1015],{"class":652,"line":1011},10,[650,1013,1014],{"class":936},"  network",[650,1016,941],{"class":940},[650,1018,1020,1023],{"class":652,"line":1019},11,[650,1021,1022],{"class":936},"    allow",[650,1024,941],{"class":940},[650,1026,1028,1031],{"class":652,"line":1027},12,[650,1029,1030],{"class":940},"      - ",[650,1032,1033],{"class":660},"\"api.anthropic.com\"\n",[650,1035,1037,1039],{"class":652,"line":1036},13,[650,1038,1030],{"class":940},[650,1040,1041],{"class":660},"\"api.openai.com\"\n",[650,1043,1045,1048,1051,1054],{"class":652,"line":1044},14,[650,1046,1047],{"class":936},"    deny",[650,1049,1050],{"class":940},": [",[650,1052,1053],{"class":660},"\"*\"",[650,1055,1056],{"class":940},"]\n",[650,1058,1060,1063],{"class":652,"line":1059},15,[650,1061,1062],{"class":936},"  filesystem",[650,1064,941],{"class":940},[650,1066,1068,1071,1073,1076],{"class":652,"line":1067},16,[650,1069,1070],{"class":936},"    read",[650,1072,1050],{"class":940},[650,1074,1075],{"class":660},"\"/data/knowledge-base\"",[650,1077,1056],{"class":940},[650,1079,1081,1084,1086,1089],{"class":652,"line":1080},17,[650,1082,1083],{"class":936},"    write",[650,1085,1050],{"class":940},[650,1087,1088],{"class":660},"\"/tmp/openclaw\"",[650,1090,1056],{"class":940},[650,1092,1094],{"class":652,"line":1093},18,[650,1095,989],{"emptyLinePlaceholder":510},[650,1097,1099,1102],{"class":652,"line":1098},19,[650,1100,1101],{"class":936},"audit",[650,1103,941],{"class":940},[650,1105,1107,1110,1112],{"class":652,"line":1106},20,[650,1108,1109],{"class":936},"  level",[650,1111,950],{"class":940},[650,1113,1114],{"class":660},"full\n",[650,1116,1118,1121,1123],{"class":652,"line":1117},21,[650,1119,1120],{"class":936},"  retention",[650,1122,950],{"class":940},[650,1124,1125],{"class":660},"90d\n",[650,1127,1129,1132,1134],{"class":652,"line":1128},22,[650,1130,1131],{"class":936},"  export",[650,1133,950],{"class":940},[650,1135,1136],{"class":660},"siem\n",[12,1138,1139],{},"Policies are version-controlled, auditable, and can be templated across multiple deployments. Changes to the policy file are validated and applied without restarting the OpenClaw gateway.",[44,1141,1143],{"id":1142},"hardware-requirements","Hardware Requirements",[12,1145,1146],{},"NemoClaw's local inference features require an NVIDIA GPU. The Privacy Router and OpenShell Runtime work on any hardware, but local Nemotron models need:",[795,1148,1149,1163],{},[798,1150,1151],{},[801,1152,1153,1156,1159,1161],{},[804,1154,1155],{},"Hardware",[804,1157,1158],{},"Local Inference",[804,1160,680],{},[804,1162,636],{},[817,1164,1165,1179,1192,1205,1218],{},[801,1166,1167,1171,1174,1177],{},[822,1168,1169],{},[28,1170,832],{},[822,1172,1173],{},"Nemotron-Mini (INT4)",[822,1175,1176],{},"Yes",[822,1178,1176],{},[801,1180,1181,1185,1188,1190],{},[822,1182,1183],{},[28,1184,848],{},[822,1186,1187],{},"Nemotron-22B (INT8)",[822,1189,1176],{},[822,1191,1176],{},[801,1193,1194,1198,1201,1203],{},[822,1195,1196],{},[28,1197,864],{},[822,1199,1200],{},"Nemotron-49B (FP8)",[822,1202,1176],{},[822,1204,1176],{},[801,1206,1207,1211,1214,1216],{},[822,1208,1209],{},[28,1210,880],{},[822,1212,1213],{},"Full Nemotron-70B",[822,1215,1176],{},[822,1217,1176],{},[801,1219,1220,1225,1228,1230],{},[822,1221,1222],{},[28,1223,1224],{},"No NVIDIA GPU",[822,1226,1227],{},"Cloud-only routing",[822,1229,1176],{},[822,1231,1176],{},[12,1233,1234],{},"Without a supported GPU, NemoClaw still provides the Privacy Router (PII stripping before cloud requests) and OpenShell sandboxing, just without local inference.",[44,1236,1238],{"id":1237},"comparison-with-other-solutions","Comparison with Other Solutions",[12,1240,1241],{},"NemoClaw isn't the only tool addressing AI security and privacy. Here's how it compares to other solutions in the space:",[795,1243,1244,1264],{},[798,1245,1246],{},[801,1247,1248,1251,1253,1256,1258,1261],{},[804,1249,1250],{},"Feature",[804,1252,532],{},[804,1254,1255],{},"Azure AI Content Safety",[804,1257,559],{},[804,1259,1260],{},"Presidio",[804,1262,1263],{},"Custom Proxy",[817,1265,1266,1287,1304,1323,1341,1363,1382,1402],{},[801,1267,1268,1273,1276,1279,1282,1284],{},[822,1269,1270],{},[28,1271,1272],{},"PII detection/redaction",[822,1274,1275],{},"Yes (local)",[822,1277,1278],{},"Yes (cloud)",[822,1280,1281],{},"No",[822,1283,1275],{},[822,1285,1286],{},"Manual",[801,1288,1289,1294,1296,1298,1300,1302],{},[822,1290,1291],{},[28,1292,1293],{},"Hybrid routing (local/cloud)",[822,1295,1176],{},[822,1297,1281],{},[822,1299,1281],{},[822,1301,1281],{},[822,1303,1286],{},[801,1305,1306,1311,1314,1316,1318,1320],{},[822,1307,1308],{},[28,1309,1310],{},"Sandboxed execution",[822,1312,1313],{},"Yes (OpenShell)",[822,1315,1281],{},[822,1317,1281],{},[822,1319,1281],{},[822,1321,1322],{},"Varies",[801,1324,1325,1330,1333,1335,1337,1339],{},[822,1326,1327],{},[28,1328,1329],{},"Local inference",[822,1331,1332],{},"Yes (Nemotron)",[822,1334,1281],{},[822,1336,1281],{},[822,1338,1281],{},[822,1340,1322],{},[801,1342,1343,1348,1351,1354,1357,1360],{},[822,1344,1345],{},[28,1346,1347],{},"Declarative policies",[822,1349,1350],{},"Yes (YAML)",[822,1352,1353],{},"Portal/API",[822,1355,1356],{},"Yes (Colang)",[822,1358,1359],{},"Config",[822,1361,1362],{},"Custom",[801,1364,1365,1370,1373,1376,1378,1380],{},[822,1366,1367],{},[28,1368,1369],{},"OpenClaw integration",[822,1371,1372],{},"Native",[822,1374,1375],{},"Requires adapter",[822,1377,1375],{},[822,1379,1375],{},[822,1381,1362],{},[801,1383,1384,1389,1392,1394,1396,1399],{},[822,1385,1386],{},[28,1387,1388],{},"Open source",[822,1390,1391],{},"Yes (Apache 2.0)",[822,1393,1281],{},[822,1395,1391],{},[822,1397,1398],{},"Yes (MIT)",[822,1400,1401],{},"N/A",[801,1403,1404,1408,1411,1413,1416,1418],{},[822,1405,1406],{},[28,1407,744],{},[822,1409,1410],{},"Built-in",[822,1412,1410],{},[822,1414,1415],{},"Limited",[822,1417,1281],{},[822,1419,1362],{},[12,1421,1422],{},"Each tool has its strengths. Azure AI Content Safety offers managed scalability, NeMo Guardrails provides fine-grained conversational control, and Presidio excels at standalone PII detection across multiple languages. NemoClaw's differentiator is that it combines privacy routing, sandboxing, and local inference into a single stack purpose-built for the OpenClaw ecosystem.",[44,1424,1426],{"id":1425},"why-this-matters-for-openclaw-users","Why This Matters for OpenClaw Users",[12,1428,1429,1430,1433],{},"OpenClaw's biggest barrier to enterprise adoption has been the ",[28,1431,1432],{},"security question",": how do you give an AI assistant access to your files, emails, and tools without risking data leaks? NemoClaw directly answers that concern.",[12,1435,1436],{},"As Jensen Huang put it during the GTC keynote:",[1438,1439,1440],"blockquote",{},[12,1441,1442],{},"\"AI assistants need to be as secure as they are capable. NemoClaw makes OpenClaw enterprise-ready. Privacy by default, not privacy as an afterthought.\"",[12,1444,1445],{},"Key implications:",[112,1447,1448,1454,1460,1466],{},[115,1449,1450,1453],{},[28,1451,1452],{},"Regulated industries"," (healthcare, finance, legal) can now deploy OpenClaw with auditable security controls.",[115,1455,1456,1459],{},[28,1457,1458],{},"Enterprise IT teams"," get the policy-based governance they need to approve OpenClaw deployments.",[115,1461,1462,1465],{},[28,1463,1464],{},"Open-source ecosystem"," benefits from NVIDIA's investment in hardened tooling.",[115,1467,1468,1471],{},[28,1469,1470],{},"Self-hosters"," get a plug-and-play security layer without building custom solutions.",[12,1473,1474],{},"The community response has been significant. Within 24 hours of the announcement, the NemoClaw repository received over 2,000 GitHub stars. Discussions in the OpenClaw community are already focused on integration patterns, particularly around combining NemoClaw's Privacy Router with existing OpenClaw skill ecosystems. Several contributors have begun working on adapters for non-Nemotron models, signaling that the architecture's modular design is resonating with developers.",[44,1476,1478],{"id":1477},"whats-next-for-nemoclaw","What's Next for NemoClaw",[12,1480,1481],{},"NemoClaw is currently in alpha, but NVIDIA has outlined a clear path forward:",[112,1483,1484,1490,1496],{},[115,1485,1486,1489],{},[28,1487,1488],{},"NemoClaw 1.0 (expected Q3 2026)"," will bring stable API surfaces, production-ready OpenShell with SELinux and AppArmor profiles, and compliance policy templates for SOC 2, HIPAA, and GDPR. Based on NVIDIA's announced roadmap, the 1.0 release is expected to mark the transition from experimental to production-supported.",[115,1491,1492,1495],{},[28,1493,1494],{},"Multi-model support"," is expected in the 1.0 release, covering GGUF and TensorRT-LLM-compatible models beyond the Nemotron family. This aligns with NVIDIA's stated goal of making NemoClaw model-agnostic for local inference.",[115,1497,1498,1501],{},[28,1499,1500],{},"NVIDIA AI Enterprise integration"," is a logical next step, bringing commercial support and SLAs through the AI Enterprise platform, including NGC model lifecycle management for organizations that need vendor-backed guarantees.",[12,1503,1504],{},"It's worth noting that these are forward-looking items based on NVIDIA's public communications. The open-source nature of the project means community contributions will likely shape the roadmap alongside NVIDIA's plans.",[44,1506,1508],{"id":1507},"what-this-means-for-clawnest-users","What This Means for ClawNest Users",[12,1510,1511],{},"At ClawNest, we're closely tracking NemoClaw's development. Here's how it fits into our managed hosting platform:",[112,1513,1514,1520,1526],{},[115,1515,1516,1519],{},[28,1517,1518],{},"Complementary, not competing."," NemoClaw handles security policy and local inference; ClawNest handles uptime, backups, and deployment simplicity. The two work well together.",[115,1521,1522,1525],{},[28,1523,1524],{},"Privacy Router compatibility."," ClawNest instances can be configured to route through a NemoClaw Privacy Router running on your own network, giving you PII stripping before data reaches our servers.",[115,1527,1528,1531],{},[28,1529,1530],{},"Enterprise plans."," We're evaluating NemoClaw integration for future ClawNest enterprise tiers, including managed OpenShell sandboxing and policy templates.",[12,1533,1534],{},"If you're running OpenClaw on ClawNest today and want to layer NemoClaw on top for additional security controls, the Privacy Router can sit between your messaging channels and your ClawNest instance as a local proxy.",[44,1536,1538],{"id":1537},"getting-started-with-nemoclaw","Getting Started with NemoClaw",[12,1540,1541,1542,1545],{},"NemoClaw is available now in ",[28,1543,1544],{},"alpha"," on GitHub:",[1547,1548,1549,1555,1563,1571],"ol",{},[115,1550,1551,1554],{},[28,1552,1553],{},"Requirements",": Node.js 22+, OpenClaw 0.9+, NVIDIA GPU (optional, for local inference)",[115,1556,1557,950,1560],{},[28,1558,1559],{},"Install",[231,1561,1562],{},"npm install -g nemoclaw@alpha",[115,1564,1565,950,1568,1570],{},[28,1566,1567],{},"Initialize",[231,1569,580],{}," to generate your first policy blueprint",[115,1572,1573,1576],{},[28,1574,1575],{},"Connect",": point NemoClaw at your existing OpenClaw gateway",[12,1578,1579],{},"Note that NemoClaw is in early alpha. Expect breaking changes and limited model support. NVIDIA has committed to a stable 1.0 release by Q3 2026.",[12,1581,1582,1583,157],{},"For the full documentation and source code, visit the ",[1584,1585,1589],"a",{"href":1586,"rel":1587},"https://github.com/NVIDIA/NemoClaw",[1588],"nofollow","NemoClaw GitHub repository",[1591,1592],"hr",{},[12,1594,1595],{},[1596,1597,1598,1599,1602,1603,1607],"em",{},"Want to run OpenClaw without the complexity of self-hosting? ",[1584,1600,295],{"href":1601},"/openclaw-hosting"," provides fully managed OpenClaw instances with automatic backups, a web dashboard, and deployment in minutes. ",[1584,1604,1606],{"href":1605},"/pricing","Start your free trial"," today.",[1609,1610,1611],"style",{},"html pre.shiki code .sScJk, html code.shiki .sScJk{--shiki-default:#6F42C1;--shiki-dark:#B392F0}html pre.shiki code .sZZnC, html code.shiki .sZZnC{--shiki-default:#032F62;--shiki-dark:#9ECBFF}html pre.shiki code .sj4cs, html code.shiki .sj4cs{--shiki-default:#005CC5;--shiki-dark:#79B8FF}html .default .shiki span {color: var(--shiki-default);background: var(--shiki-default-bg);font-style: var(--shiki-default-font-style);font-weight: var(--shiki-default-font-weight);text-decoration: var(--shiki-default-text-decoration);}html .shiki span {color: var(--shiki-default);background: var(--shiki-default-bg);font-style: var(--shiki-default-font-style);font-weight: var(--shiki-default-font-weight);text-decoration: var(--shiki-default-text-decoration);}html .dark .shiki span {color: var(--shiki-dark);background: var(--shiki-dark-bg);font-style: var(--shiki-dark-font-style);font-weight: var(--shiki-dark-font-weight);text-decoration: var(--shiki-dark-text-decoration);}html.dark .shiki span {color: var(--shiki-dark);background: var(--shiki-dark-bg);font-style: var(--shiki-dark-font-style);font-weight: var(--shiki-dark-font-weight);text-decoration: var(--shiki-dark-text-decoration);}html pre.shiki code .sJ8bj, html code.shiki .sJ8bj{--shiki-default:#6A737D;--shiki-dark:#6A737D}html pre.shiki code .s9eBZ, html code.shiki .s9eBZ{--shiki-default:#22863A;--shiki-dark:#85E89D}html pre.shiki code .sVt8B, html code.shiki .sVt8B{--shiki-default:#24292E;--shiki-dark:#E1E4E8}",{"title":489,"searchDepth":490,"depth":490,"links":1613},[1614,1615,1616,1617,1618,1624,1625,1626,1627,1628,1629],{"id":525,"depth":490,"text":526},{"id":542,"depth":490,"text":543},{"id":584,"depth":490,"text":585},{"id":625,"depth":490,"text":626},{"id":674,"depth":490,"text":675,"children":1619},[1620,1621,1622,1623],{"id":679,"depth":944,"text":680},{"id":717,"depth":944,"text":718},{"id":757,"depth":944,"text":758},{"id":910,"depth":944,"text":911},{"id":1142,"depth":490,"text":1143},{"id":1237,"depth":490,"text":1238},{"id":1425,"depth":490,"text":1426},{"id":1477,"depth":490,"text":1478},{"id":1507,"depth":490,"text":1508},{"id":1537,"depth":490,"text":1538},"2026-03-17","Deep dive into NVIDIA's open-source NemoClaw security stack for OpenClaw. Covers privacy routing, kernel-level sandboxing, Nemotron local inference, and what the GTC 2026 announcement means for the ecosystem.","/blog/nanoclaw_openclaw_clawnest.png",{},"/blog/nvidia-nemoclaw-openclaw-security",{"title":520,"description":1631},"blog/nvidia-nemoclaw-openclaw-security",[68,34,532,1638,1639,1640,632,1641],"Security","Enterprise","GTC 2026","Privacy","3gt4codd_6LFM3F2oTE13_whQIT4uK86oP6E7dHCQbU",{"id":1644,"title":1645,"author":7,"body":1646,"date":3249,"description":3250,"draft":506,"extension":507,"image":3251,"meta":3252,"navigation":510,"path":3253,"seo":3254,"stem":3255,"tags":3256,"__hash__":3260},"blog/blog/openclaw-ollama-local-models-guide.md","How to Cut OpenClaw API Costs with Ollama and Local Models",{"type":9,"value":1647,"toc":3224},[1648,1652,1659,1662,1677,1681,1688,1693,1725,1737,1741,1749,1753,1756,1779,1782,1798,1802,1813,2044,2047,2108,2112,2118,2163,2166,2170,2176,2250,2254,2261,2281,2284,2288,2296,2300,2320,2342,2350,2541,2548,2552,2560,2573,2577,2892,2899,2903,2906,2910,2913,2927,2932,3019,3022,3026,3029,3102,3109,3113,3116,3148,3151,3155,3198,3207,3209,3221],[44,1649,1651],{"id":1650},"the-cost-problem","The Cost Problem",[12,1653,1654,1655,1658],{},"If you're self-hosting OpenClaw with cloud APIs like OpenAI or Gemini, you've probably noticed the bills adding up fast. We've seen users reporting ",[28,1656,1657],{},"$20+ per week"," on basic search and browsing tasks — and that's with moderate usage.",[12,1660,1661],{},"The reason is token volume. OpenClaw's tool-use architecture means a single web search can consume millions of tokens. The agent reasons about which tools to call, processes search results, summarizes content, and formats a response. Each step burns tokens. Multiply that by dozens of queries per day and you're looking at serious monthly spend.",[12,1663,1664,1665,1668,1669,1672,1673,1676],{},"The good news: you have options. OpenClaw supports custom model providers, which means you can run ",[28,1666,1667],{},"local models for free",", use ",[28,1670,1671],{},"cheaper cloud alternatives",", or set up a ",[28,1674,1675],{},"hybrid approach"," that gives you the best of both worlds.",[44,1678,1680],{"id":1679},"important-minimum-model-size","Important: Minimum Model Size",[12,1682,1683,1684,1687],{},"Before you get started, a critical caveat: ",[28,1685,1686],{},"OpenClaw requires models with strong tool-calling and reasoning capabilities",". Its agentic architecture involves multi-step planning, tool selection, structured output parsing, and long context windows. Small models simply can't handle this reliably.",[12,1689,1690],{},[28,1691,1692],{},"Minimum recommended sizes:",[112,1694,1695,1701,1707],{},[115,1696,1697,1700],{},[28,1698,1699],{},"32B+ parameters"," — the practical minimum for reliable OpenClaw usage (e.g. DeepSeek R1 32B, Qwen 2.5 32B)",[115,1702,1703,1706],{},[28,1704,1705],{},"14B parameters"," — may work for simple tasks but will frequently fail on multi-step workflows",[115,1708,1709,1712,1713,1716,1717,1720,1721,1724],{},[28,1710,1711],{},"7–8B parameters"," — ",[28,1714,1715],{},"not recommended",". Models like ",[231,1718,1719],{},"llama3.1:8b"," or ",[231,1722,1723],{},"mistral:7b"," lack the reasoning depth for OpenClaw's tool-use chains and will produce errors, hallucinated tool calls, or get stuck in loops",[12,1726,1727,1728,1732,1733,1736],{},"If you don't have the hardware to run 32B+ models locally, consider ",[1584,1729,1731],{"href":1730},"#option-2-use-cheaper-cloud-models-via-openrouter","OpenRouter"," or the ",[1584,1734,1675],{"href":1735},"#option-3-the-hybrid-approach"," instead.",[44,1738,1740],{"id":1739},"option-1-run-local-models-with-ollama","Option 1: Run Local Models with Ollama",[12,1742,1743,1748],{},[1584,1744,1747],{"href":1745,"rel":1746},"https://ollama.com",[1588],"Ollama"," is an open-source tool that lets you run large language models locally on your own hardware. It exposes an OpenAI-compatible API, which means OpenClaw can use it as a drop-in replacement for cloud models.",[677,1750,1752],{"id":1751},"installing-ollama","Installing Ollama",[12,1754,1755],{},"Install Ollama on the same machine as OpenClaw (or a separate machine — more on that later):",[642,1757,1759],{"className":644,"code":1758,"language":646,"meta":489,"style":489},"curl -fsSL https://ollama.com/install.sh | sh\n",[231,1760,1761],{"__ignoreMap":489},[650,1762,1763,1766,1769,1772,1776],{"class":652,"line":653},[650,1764,1765],{"class":656},"curl",[650,1767,1768],{"class":664}," -fsSL",[650,1770,1771],{"class":660}," https://ollama.com/install.sh",[650,1773,1775],{"class":1774},"szBVR"," |",[650,1777,1778],{"class":656}," sh\n",[12,1780,1781],{},"Then pull a model. We recommend starting with a 32B+ model:",[642,1783,1785],{"className":644,"code":1784,"language":646,"meta":489,"style":489},"ollama pull qwen3:32b\n",[231,1786,1787],{"__ignoreMap":489},[650,1788,1789,1792,1795],{"class":652,"line":653},[650,1790,1791],{"class":656},"ollama",[650,1793,1794],{"class":660}," pull",[650,1796,1797],{"class":660}," qwen3:32b\n",[677,1799,1801],{"id":1800},"configuring-openclaw-to-use-ollama","Configuring OpenClaw to Use Ollama",[12,1803,1804,1805,1808,1809,1812],{},"In your ",[231,1806,1807],{},"openclaw.json"," config, add Ollama as a custom provider under ",[231,1810,1811],{},"models.providers",":",[642,1814,1818],{"className":1815,"code":1816,"language":1817,"meta":489,"style":489},"language-json shiki shiki-themes github-light github-dark","{\n  \"models\": {\n    \"providers\": {\n      \"ollama\": {\n        \"baseUrl\": \"http://localhost:11434/v1\",\n        \"api\": \"openai-completions\",\n        \"apiKey\": \"ollama\",\n        \"models\": [\n          {\n            \"id\": \"qwen3:32b\",\n            \"name\": \"Qwen 3 32B\",\n            \"reasoning\": true,\n            \"input\": [\"text\"],\n            \"cost\": { \"input\": 0, \"output\": 0, \"cacheRead\": 0, \"cacheWrite\": 0 },\n            \"contextWindow\": 32768,\n            \"maxTokens\": 8192\n          }\n        ]\n      }\n    }\n  }\n}\n","json",[231,1819,1820,1825,1833,1840,1847,1860,1872,1884,1892,1897,1909,1921,1933,1946,1992,2004,2014,2019,2024,2029,2034,2039],{"__ignoreMap":489},[650,1821,1822],{"class":652,"line":653},[650,1823,1824],{"class":940},"{\n",[650,1826,1827,1830],{"class":652,"line":490},[650,1828,1829],{"class":664},"  \"models\"",[650,1831,1832],{"class":940},": {\n",[650,1834,1835,1838],{"class":652,"line":944},[650,1836,1837],{"class":664},"    \"providers\"",[650,1839,1832],{"class":940},[650,1841,1842,1845],{"class":652,"line":956},[650,1843,1844],{"class":664},"      \"ollama\"",[650,1846,1832],{"class":940},[650,1848,1849,1852,1854,1857],{"class":652,"line":964},[650,1850,1851],{"class":664},"        \"baseUrl\"",[650,1853,950],{"class":940},[650,1855,1856],{"class":660},"\"http://localhost:11434/v1\"",[650,1858,1859],{"class":940},",\n",[650,1861,1862,1865,1867,1870],{"class":652,"line":975},[650,1863,1864],{"class":664},"        \"api\"",[650,1866,950],{"class":940},[650,1868,1869],{"class":660},"\"openai-completions\"",[650,1871,1859],{"class":940},[650,1873,1874,1877,1879,1882],{"class":652,"line":986},[650,1875,1876],{"class":664},"        \"apiKey\"",[650,1878,950],{"class":940},[650,1880,1881],{"class":660},"\"ollama\"",[650,1883,1859],{"class":940},[650,1885,1886,1889],{"class":652,"line":992},[650,1887,1888],{"class":664},"        \"models\"",[650,1890,1891],{"class":940},": [\n",[650,1893,1894],{"class":652,"line":1000},[650,1895,1896],{"class":940},"          {\n",[650,1898,1899,1902,1904,1907],{"class":652,"line":1011},[650,1900,1901],{"class":664},"            \"id\"",[650,1903,950],{"class":940},[650,1905,1906],{"class":660},"\"qwen3:32b\"",[650,1908,1859],{"class":940},[650,1910,1911,1914,1916,1919],{"class":652,"line":1019},[650,1912,1913],{"class":664},"            \"name\"",[650,1915,950],{"class":940},[650,1917,1918],{"class":660},"\"Qwen 3 32B\"",[650,1920,1859],{"class":940},[650,1922,1923,1926,1928,1931],{"class":652,"line":1027},[650,1924,1925],{"class":664},"            \"reasoning\"",[650,1927,950],{"class":940},[650,1929,1930],{"class":664},"true",[650,1932,1859],{"class":940},[650,1934,1935,1938,1940,1943],{"class":652,"line":1036},[650,1936,1937],{"class":664},"            \"input\"",[650,1939,1050],{"class":940},[650,1941,1942],{"class":660},"\"text\"",[650,1944,1945],{"class":940},"],\n",[650,1947,1948,1951,1954,1957,1959,1962,1964,1967,1969,1971,1973,1976,1978,1980,1982,1985,1987,1989],{"class":652,"line":1044},[650,1949,1950],{"class":664},"            \"cost\"",[650,1952,1953],{"class":940},": { ",[650,1955,1956],{"class":664},"\"input\"",[650,1958,950],{"class":940},[650,1960,1961],{"class":664},"0",[650,1963,35],{"class":940},[650,1965,1966],{"class":664},"\"output\"",[650,1968,950],{"class":940},[650,1970,1961],{"class":664},[650,1972,35],{"class":940},[650,1974,1975],{"class":664},"\"cacheRead\"",[650,1977,950],{"class":940},[650,1979,1961],{"class":664},[650,1981,35],{"class":940},[650,1983,1984],{"class":664},"\"cacheWrite\"",[650,1986,950],{"class":940},[650,1988,1961],{"class":664},[650,1990,1991],{"class":940}," },\n",[650,1993,1994,1997,1999,2002],{"class":652,"line":1059},[650,1995,1996],{"class":664},"            \"contextWindow\"",[650,1998,950],{"class":940},[650,2000,2001],{"class":664},"32768",[650,2003,1859],{"class":940},[650,2005,2006,2009,2011],{"class":652,"line":1067},[650,2007,2008],{"class":664},"            \"maxTokens\"",[650,2010,950],{"class":940},[650,2012,2013],{"class":664},"8192\n",[650,2015,2016],{"class":652,"line":1080},[650,2017,2018],{"class":940},"          }\n",[650,2020,2021],{"class":652,"line":1093},[650,2022,2023],{"class":940},"        ]\n",[650,2025,2026],{"class":652,"line":1098},[650,2027,2028],{"class":940},"      }\n",[650,2030,2031],{"class":652,"line":1106},[650,2032,2033],{"class":940},"    }\n",[650,2035,2036],{"class":652,"line":1117},[650,2037,2038],{"class":940},"  }\n",[650,2040,2041],{"class":652,"line":1128},[650,2042,2043],{"class":940},"}\n",[12,2045,2046],{},"The key settings:",[112,2048,2049,2059,2068,2076],{},[115,2050,2051,2054,2055,2058],{},[28,2052,2053],{},"baseUrl"," — points to Ollama's OpenAI-compatible endpoint (port 11434 by default, with ",[231,2056,2057],{},"/v1"," path)",[115,2060,2061,2064,2065,2067],{},[28,2062,2063],{},"api"," — must be ",[231,2066,1869],{}," (the OpenAI Chat Completions API adapter)",[115,2069,2070,2073,2074],{},[28,2071,2072],{},"apiKey"," — required by OpenClaw even though Ollama doesn't need authentication. Use any placeholder value like ",[231,2075,1881],{},[115,2077,2078,2081,2082,2085,2086,35,2089,35,2092,35,2095,35,2098,35,2101,2104,2105],{},[28,2079,2080],{},"models"," — list the models you've pulled with ",[231,2083,2084],{},"ollama pull",". Each model entry requires ",[231,2087,2088],{},"id",[231,2090,2091],{},"name",[231,2093,2094],{},"reasoning",[231,2096,2097],{},"input",[231,2099,2100],{},"cost",[231,2102,2103],{},"contextWindow",", and ",[231,2106,2107],{},"maxTokens",[677,2109,2111],{"id":2110},"setting-a-local-model-as-default","Setting a Local Model as Default",[12,2113,2114,2115,1812],{},"To make OpenClaw use your local model by default instead of a cloud API, configure ",[231,2116,2117],{},"agents.defaults",[642,2119,2121],{"className":1815,"code":2120,"language":1817,"meta":489,"style":489},"{\n  \"agents\": {\n    \"defaults\": {\n      \"model\": \"ollama/qwen3:32b\"\n    }\n  }\n}\n",[231,2122,2123,2127,2134,2141,2151,2155,2159],{"__ignoreMap":489},[650,2124,2125],{"class":652,"line":653},[650,2126,1824],{"class":940},[650,2128,2129,2132],{"class":652,"line":490},[650,2130,2131],{"class":664},"  \"agents\"",[650,2133,1832],{"class":940},[650,2135,2136,2139],{"class":652,"line":944},[650,2137,2138],{"class":664},"    \"defaults\"",[650,2140,1832],{"class":940},[650,2142,2143,2146,2148],{"class":652,"line":956},[650,2144,2145],{"class":664},"      \"model\"",[650,2147,950],{"class":940},[650,2149,2150],{"class":660},"\"ollama/qwen3:32b\"\n",[650,2152,2153],{"class":652,"line":964},[650,2154,2033],{"class":940},[650,2156,2157],{"class":652,"line":975},[650,2158,2038],{"class":940},[650,2160,2161],{"class":652,"line":986},[650,2162,2043],{"class":940},[12,2164,2165],{},"Now every new conversation starts with your local model — zero API cost.",[44,2167,2169],{"id":2168},"hardware-requirements-for-local-models","Hardware Requirements for Local Models",[12,2171,2172,2173,2175],{},"Local models run on your CPU or GPU. The limiting factor is almost always ",[28,2174,433],{}," — the model needs to fit entirely in RAM (or VRAM for GPU inference).",[795,2177,2178,2194],{},[798,2179,2180],{},[801,2181,2182,2185,2188,2191],{},[804,2183,2184],{},"RAM / VRAM",[804,2186,2187],{},"Model Size",[804,2189,2190],{},"OpenClaw Compatibility",[804,2192,2193],{},"Examples",[817,2195,2196,2210,2223,2236],{},[801,2197,2198,2201,2204,2207],{},[822,2199,2200],{},"8 GB",[822,2202,2203],{},"7B parameters",[822,2205,2206],{},"Not compatible — too small for tool-use",[822,2208,2209],{},"Llama 3.1 8B, Mistral 7B",[801,2211,2212,2215,2217,2220],{},[822,2213,2214],{},"16 GB",[822,2216,1705],{},[822,2218,2219],{},"Limited — simple tasks only",[822,2221,2222],{},"Qwen 2.5 14B",[801,2224,2225,2228,2230,2233],{},[822,2226,2227],{},"32 GB+",[822,2229,1699],{},[822,2231,2232],{},"Recommended — reliable for most tasks",[822,2234,2235],{},"DeepSeek R1 32B, Qwen 3 32B",[801,2237,2238,2241,2244,2247],{},[822,2239,2240],{},"48 GB+ VRAM (GPU)",[822,2242,2243],{},"70B+ parameters",[822,2245,2246],{},"Excellent — comparable to mid-tier cloud models",[822,2248,2249],{},"Llama 3.1 70B, Mixtral 8x22B",[677,2251,2253],{"id":2252},"gpu-recommendations","GPU Recommendations",[12,2255,2256,2257,2260],{},"Running models on a GPU is ",[28,2258,2259],{},"5–10x faster"," than CPU inference. If you're serious about local models:",[112,2262,2263,2269,2275],{},[115,2264,2265,2268],{},[28,2266,2267],{},"RTX 3090 (24 GB VRAM)"," — runs 14B models at full speed, 32B models with quantization",[115,2270,2271,2274],{},[28,2272,2273],{},"RTX 4090 (24 GB VRAM)"," — same capacity, faster inference",[115,2276,2277,2280],{},[28,2278,2279],{},"Dual GPUs or 48 GB+ VRAM"," — needed for 70B+ models without heavy quantization",[12,2282,2283],{},"CPU-only inference works but expect slower response times — around 5–15 tokens per second depending on your hardware and model size.",[44,2285,2287],{"id":2286},"running-ollama-on-a-separate-machine","Running Ollama on a Separate Machine",[12,2289,2290,2291,157],{},"Your VPS probably doesn't have a GPU. But your desktop at home might. You can run Ollama on a home machine with a GPU and connect your VPS to it securely using ",[1584,2292,2295],{"href":2293,"rel":2294},"https://tailscale.com",[1588],"Tailscale",[677,2297,2299],{"id":2298},"setup","Setup",[1547,2301,2302,2308,2314],{},[115,2303,2304,2307],{},[28,2305,2306],{},"Install Tailscale"," on both your VPS and your home machine",[115,2309,2310,2313],{},[28,2311,2312],{},"Install Ollama"," on your home machine (the one with the GPU)",[115,2315,2316,2319],{},[28,2317,2318],{},"Start Ollama"," with network access enabled:",[642,2321,2323],{"className":644,"code":2322,"language":646,"meta":489,"style":489},"OLLAMA_HOST=0.0.0.0 ollama serve\n",[231,2324,2325],{"__ignoreMap":489},[650,2326,2327,2330,2333,2336,2339],{"class":652,"line":653},[650,2328,2329],{"class":940},"OLLAMA_HOST",[650,2331,2332],{"class":1774},"=",[650,2334,2335],{"class":660},"0.0.0.0",[650,2337,2338],{"class":656}," ollama",[650,2340,2341],{"class":660}," serve\n",[1547,2343,2344],{"start":956},[115,2345,2346,2349],{},[28,2347,2348],{},"Update OpenClaw config"," on your VPS to point to the Tailscale IP:",[642,2351,2353],{"className":1815,"code":2352,"language":1817,"meta":489,"style":489},"{\n  \"models\": {\n    \"providers\": {\n      \"ollama\": {\n        \"baseUrl\": \"http://100.x.x.x:11434/v1\",\n        \"api\": \"openai-completions\",\n        \"apiKey\": \"ollama\",\n        \"models\": [\n          {\n            \"id\": \"deepseek-r1:32b\",\n            \"name\": \"DeepSeek R1 32B\",\n            \"reasoning\": true,\n            \"input\": [\"text\"],\n            \"cost\": { \"input\": 0, \"output\": 0, \"cacheRead\": 0, \"cacheWrite\": 0 },\n            \"contextWindow\": 65536,\n            \"maxTokens\": 8192\n          }\n        ]\n      }\n    }\n  }\n}\n",[231,2354,2355,2359,2365,2371,2377,2388,2398,2408,2414,2418,2429,2440,2450,2460,2498,2509,2517,2521,2525,2529,2533,2537],{"__ignoreMap":489},[650,2356,2357],{"class":652,"line":653},[650,2358,1824],{"class":940},[650,2360,2361,2363],{"class":652,"line":490},[650,2362,1829],{"class":664},[650,2364,1832],{"class":940},[650,2366,2367,2369],{"class":652,"line":944},[650,2368,1837],{"class":664},[650,2370,1832],{"class":940},[650,2372,2373,2375],{"class":652,"line":956},[650,2374,1844],{"class":664},[650,2376,1832],{"class":940},[650,2378,2379,2381,2383,2386],{"class":652,"line":964},[650,2380,1851],{"class":664},[650,2382,950],{"class":940},[650,2384,2385],{"class":660},"\"http://100.x.x.x:11434/v1\"",[650,2387,1859],{"class":940},[650,2389,2390,2392,2394,2396],{"class":652,"line":975},[650,2391,1864],{"class":664},[650,2393,950],{"class":940},[650,2395,1869],{"class":660},[650,2397,1859],{"class":940},[650,2399,2400,2402,2404,2406],{"class":652,"line":986},[650,2401,1876],{"class":664},[650,2403,950],{"class":940},[650,2405,1881],{"class":660},[650,2407,1859],{"class":940},[650,2409,2410,2412],{"class":652,"line":992},[650,2411,1888],{"class":664},[650,2413,1891],{"class":940},[650,2415,2416],{"class":652,"line":1000},[650,2417,1896],{"class":940},[650,2419,2420,2422,2424,2427],{"class":652,"line":1011},[650,2421,1901],{"class":664},[650,2423,950],{"class":940},[650,2425,2426],{"class":660},"\"deepseek-r1:32b\"",[650,2428,1859],{"class":940},[650,2430,2431,2433,2435,2438],{"class":652,"line":1019},[650,2432,1913],{"class":664},[650,2434,950],{"class":940},[650,2436,2437],{"class":660},"\"DeepSeek R1 32B\"",[650,2439,1859],{"class":940},[650,2441,2442,2444,2446,2448],{"class":652,"line":1027},[650,2443,1925],{"class":664},[650,2445,950],{"class":940},[650,2447,1930],{"class":664},[650,2449,1859],{"class":940},[650,2451,2452,2454,2456,2458],{"class":652,"line":1036},[650,2453,1937],{"class":664},[650,2455,1050],{"class":940},[650,2457,1942],{"class":660},[650,2459,1945],{"class":940},[650,2461,2462,2464,2466,2468,2470,2472,2474,2476,2478,2480,2482,2484,2486,2488,2490,2492,2494,2496],{"class":652,"line":1044},[650,2463,1950],{"class":664},[650,2465,1953],{"class":940},[650,2467,1956],{"class":664},[650,2469,950],{"class":940},[650,2471,1961],{"class":664},[650,2473,35],{"class":940},[650,2475,1966],{"class":664},[650,2477,950],{"class":940},[650,2479,1961],{"class":664},[650,2481,35],{"class":940},[650,2483,1975],{"class":664},[650,2485,950],{"class":940},[650,2487,1961],{"class":664},[650,2489,35],{"class":940},[650,2491,1984],{"class":664},[650,2493,950],{"class":940},[650,2495,1961],{"class":664},[650,2497,1991],{"class":940},[650,2499,2500,2502,2504,2507],{"class":652,"line":1059},[650,2501,1996],{"class":664},[650,2503,950],{"class":940},[650,2505,2506],{"class":664},"65536",[650,2508,1859],{"class":940},[650,2510,2511,2513,2515],{"class":652,"line":1067},[650,2512,2008],{"class":664},[650,2514,950],{"class":940},[650,2516,2013],{"class":664},[650,2518,2519],{"class":652,"line":1080},[650,2520,2018],{"class":940},[650,2522,2523],{"class":652,"line":1093},[650,2524,2023],{"class":940},[650,2526,2527],{"class":652,"line":1098},[650,2528,2028],{"class":940},[650,2530,2531],{"class":652,"line":1106},[650,2532,2033],{"class":940},[650,2534,2535],{"class":652,"line":1117},[650,2536,2038],{"class":940},[650,2538,2539],{"class":652,"line":1128},[650,2540,2043],{"class":940},[12,2542,2543,2544,2547],{},"Replace ",[231,2545,2546],{},"100.x.x.x"," with your home machine's Tailscale IP. The connection is encrypted and doesn't require opening any ports on your home network.",[44,2549,2551],{"id":2550},"option-2-use-cheaper-cloud-models-via-openrouter","Option 2: Use Cheaper Cloud Models via OpenRouter",[12,2553,2554,2555,2559],{},"Not everyone has GPU hardware at home. ",[1584,2556,1731],{"href":2557,"rel":2558},"https://openrouter.ai",[1588]," aggregates dozens of AI models and lets you pay per token — often at a fraction of the cost of direct API access.",[12,2561,2562,2563,35,2566,2104,2569,2572],{},"Models like ",[28,2564,2565],{},"Gemini Flash",[28,2567,2568],{},"Llama 3.1 70B",[28,2570,2571],{},"Mistral Large"," are available at significantly lower rates than GPT or Claude, and work well for routine OpenClaw tasks.",[677,2574,2576],{"id":2575},"openrouter-config","OpenRouter Config",[642,2578,2580],{"className":1815,"code":2579,"language":1817,"meta":489,"style":489},"{\n  \"models\": {\n    \"providers\": {\n      \"openrouter\": {\n        \"baseUrl\": \"https://openrouter.ai/api/v1\",\n        \"api\": \"openai-completions\",\n        \"apiKey\": \"sk-or-your-key-here\",\n        \"models\": [\n          {\n            \"id\": \"google/gemini-2.0-flash-001\",\n            \"name\": \"Gemini 2.0 Flash\",\n            \"reasoning\": false,\n            \"input\": [\"text\"],\n            \"cost\": { \"input\": 0.1, \"output\": 0.4, \"cacheRead\": 0.025, \"cacheWrite\": 0.1 },\n            \"contextWindow\": 1048576,\n            \"maxTokens\": 8192\n          },\n          {\n            \"id\": \"meta-llama/llama-3.1-70b-instruct\",\n            \"name\": \"Llama 3.1 70B\",\n            \"reasoning\": false,\n            \"input\": [\"text\"],\n            \"cost\": { \"input\": 0.39, \"output\": 0.39, \"cacheRead\": 0, \"cacheWrite\": 0 },\n            \"contextWindow\": 131072,\n            \"maxTokens\": 8192\n          }\n        ]\n      }\n    }\n  }\n}\n",[231,2581,2582,2586,2592,2598,2605,2616,2626,2637,2643,2647,2658,2669,2680,2690,2731,2742,2750,2755,2759,2770,2781,2791,2801,2841,2853,2862,2867,2872,2877,2882,2887],{"__ignoreMap":489},[650,2583,2584],{"class":652,"line":653},[650,2585,1824],{"class":940},[650,2587,2588,2590],{"class":652,"line":490},[650,2589,1829],{"class":664},[650,2591,1832],{"class":940},[650,2593,2594,2596],{"class":652,"line":944},[650,2595,1837],{"class":664},[650,2597,1832],{"class":940},[650,2599,2600,2603],{"class":652,"line":956},[650,2601,2602],{"class":664},"      \"openrouter\"",[650,2604,1832],{"class":940},[650,2606,2607,2609,2611,2614],{"class":652,"line":964},[650,2608,1851],{"class":664},[650,2610,950],{"class":940},[650,2612,2613],{"class":660},"\"https://openrouter.ai/api/v1\"",[650,2615,1859],{"class":940},[650,2617,2618,2620,2622,2624],{"class":652,"line":975},[650,2619,1864],{"class":664},[650,2621,950],{"class":940},[650,2623,1869],{"class":660},[650,2625,1859],{"class":940},[650,2627,2628,2630,2632,2635],{"class":652,"line":986},[650,2629,1876],{"class":664},[650,2631,950],{"class":940},[650,2633,2634],{"class":660},"\"sk-or-your-key-here\"",[650,2636,1859],{"class":940},[650,2638,2639,2641],{"class":652,"line":992},[650,2640,1888],{"class":664},[650,2642,1891],{"class":940},[650,2644,2645],{"class":652,"line":1000},[650,2646,1896],{"class":940},[650,2648,2649,2651,2653,2656],{"class":652,"line":1011},[650,2650,1901],{"class":664},[650,2652,950],{"class":940},[650,2654,2655],{"class":660},"\"google/gemini-2.0-flash-001\"",[650,2657,1859],{"class":940},[650,2659,2660,2662,2664,2667],{"class":652,"line":1019},[650,2661,1913],{"class":664},[650,2663,950],{"class":940},[650,2665,2666],{"class":660},"\"Gemini 2.0 Flash\"",[650,2668,1859],{"class":940},[650,2670,2671,2673,2675,2678],{"class":652,"line":1027},[650,2672,1925],{"class":664},[650,2674,950],{"class":940},[650,2676,2677],{"class":664},"false",[650,2679,1859],{"class":940},[650,2681,2682,2684,2686,2688],{"class":652,"line":1036},[650,2683,1937],{"class":664},[650,2685,1050],{"class":940},[650,2687,1942],{"class":660},[650,2689,1945],{"class":940},[650,2691,2692,2694,2696,2698,2700,2703,2705,2707,2709,2712,2714,2716,2718,2721,2723,2725,2727,2729],{"class":652,"line":1044},[650,2693,1950],{"class":664},[650,2695,1953],{"class":940},[650,2697,1956],{"class":664},[650,2699,950],{"class":940},[650,2701,2702],{"class":664},"0.1",[650,2704,35],{"class":940},[650,2706,1966],{"class":664},[650,2708,950],{"class":940},[650,2710,2711],{"class":664},"0.4",[650,2713,35],{"class":940},[650,2715,1975],{"class":664},[650,2717,950],{"class":940},[650,2719,2720],{"class":664},"0.025",[650,2722,35],{"class":940},[650,2724,1984],{"class":664},[650,2726,950],{"class":940},[650,2728,2702],{"class":664},[650,2730,1991],{"class":940},[650,2732,2733,2735,2737,2740],{"class":652,"line":1059},[650,2734,1996],{"class":664},[650,2736,950],{"class":940},[650,2738,2739],{"class":664},"1048576",[650,2741,1859],{"class":940},[650,2743,2744,2746,2748],{"class":652,"line":1067},[650,2745,2008],{"class":664},[650,2747,950],{"class":940},[650,2749,2013],{"class":664},[650,2751,2752],{"class":652,"line":1080},[650,2753,2754],{"class":940},"          },\n",[650,2756,2757],{"class":652,"line":1093},[650,2758,1896],{"class":940},[650,2760,2761,2763,2765,2768],{"class":652,"line":1098},[650,2762,1901],{"class":664},[650,2764,950],{"class":940},[650,2766,2767],{"class":660},"\"meta-llama/llama-3.1-70b-instruct\"",[650,2769,1859],{"class":940},[650,2771,2772,2774,2776,2779],{"class":652,"line":1106},[650,2773,1913],{"class":664},[650,2775,950],{"class":940},[650,2777,2778],{"class":660},"\"Llama 3.1 70B\"",[650,2780,1859],{"class":940},[650,2782,2783,2785,2787,2789],{"class":652,"line":1117},[650,2784,1925],{"class":664},[650,2786,950],{"class":940},[650,2788,2677],{"class":664},[650,2790,1859],{"class":940},[650,2792,2793,2795,2797,2799],{"class":652,"line":1128},[650,2794,1937],{"class":664},[650,2796,1050],{"class":940},[650,2798,1942],{"class":660},[650,2800,1945],{"class":940},[650,2802,2804,2806,2808,2810,2812,2815,2817,2819,2821,2823,2825,2827,2829,2831,2833,2835,2837,2839],{"class":652,"line":2803},23,[650,2805,1950],{"class":664},[650,2807,1953],{"class":940},[650,2809,1956],{"class":664},[650,2811,950],{"class":940},[650,2813,2814],{"class":664},"0.39",[650,2816,35],{"class":940},[650,2818,1966],{"class":664},[650,2820,950],{"class":940},[650,2822,2814],{"class":664},[650,2824,35],{"class":940},[650,2826,1975],{"class":664},[650,2828,950],{"class":940},[650,2830,1961],{"class":664},[650,2832,35],{"class":940},[650,2834,1984],{"class":664},[650,2836,950],{"class":940},[650,2838,1961],{"class":664},[650,2840,1991],{"class":940},[650,2842,2844,2846,2848,2851],{"class":652,"line":2843},24,[650,2845,1996],{"class":664},[650,2847,950],{"class":940},[650,2849,2850],{"class":664},"131072",[650,2852,1859],{"class":940},[650,2854,2856,2858,2860],{"class":652,"line":2855},25,[650,2857,2008],{"class":664},[650,2859,950],{"class":940},[650,2861,2013],{"class":664},[650,2863,2865],{"class":652,"line":2864},26,[650,2866,2018],{"class":940},[650,2868,2870],{"class":652,"line":2869},27,[650,2871,2023],{"class":940},[650,2873,2875],{"class":652,"line":2874},28,[650,2876,2028],{"class":940},[650,2878,2880],{"class":652,"line":2879},29,[650,2881,2033],{"class":940},[650,2883,2885],{"class":652,"line":2884},30,[650,2886,2038],{"class":940},[650,2888,2890],{"class":652,"line":2889},31,[650,2891,2043],{"class":940},[12,2893,2894,2895,2898],{},"OpenRouter pricing varies by model, but expect to pay ",[28,2896,2897],{},"50–90% less"," than equivalent OpenAI or Anthropic models for routine tasks.",[44,2900,2902],{"id":2901},"option-3-the-hybrid-approach","Option 3: The Hybrid Approach",[12,2904,2905],{},"The smartest strategy combines local and cloud models. Use cheap or free models for everyday tasks, and reserve expensive cloud models for when you actually need them.",[677,2907,2909],{"id":2908},"how-it-works","How It Works",[12,2911,2912],{},"OpenClaw lets you assign different models to different agents. The idea:",[112,2914,2915,2921],{},[115,2916,2917,2920],{},[28,2918,2919],{},"Routine tasks"," (web search, summarization, formatting) → Ollama local model or cheap OpenRouter model",[115,2922,2923,2926],{},[28,2924,2925],{},"Complex reasoning"," (code generation, multi-step analysis, creative writing) → GPT, Claude, or Gemini Pro",[12,2928,2929,2930,1812],{},"Configure per-agent model selection in your ",[231,2931,1807],{},[642,2933,2935],{"className":1815,"code":2934,"language":1817,"meta":489,"style":489},"{\n  \"agents\": {\n    \"defaults\": {\n      \"model\": \"ollama/qwen3:32b\"\n    },\n    \"overrides\": {\n      \"coder\": { \"model\": \"anthropic/claude-sonnet-4-5-20250929\" },\n      \"researcher\": { \"model\": \"openrouter/google/gemini-2.0-flash-001\" }\n    }\n  }\n}\n",[231,2936,2937,2941,2947,2953,2961,2966,2973,2990,3007,3011,3015],{"__ignoreMap":489},[650,2938,2939],{"class":652,"line":653},[650,2940,1824],{"class":940},[650,2942,2943,2945],{"class":652,"line":490},[650,2944,2131],{"class":664},[650,2946,1832],{"class":940},[650,2948,2949,2951],{"class":652,"line":944},[650,2950,2138],{"class":664},[650,2952,1832],{"class":940},[650,2954,2955,2957,2959],{"class":652,"line":956},[650,2956,2145],{"class":664},[650,2958,950],{"class":940},[650,2960,2150],{"class":660},[650,2962,2963],{"class":652,"line":964},[650,2964,2965],{"class":940},"    },\n",[650,2967,2968,2971],{"class":652,"line":975},[650,2969,2970],{"class":664},"    \"overrides\"",[650,2972,1832],{"class":940},[650,2974,2975,2978,2980,2983,2985,2988],{"class":652,"line":986},[650,2976,2977],{"class":664},"      \"coder\"",[650,2979,1953],{"class":940},[650,2981,2982],{"class":664},"\"model\"",[650,2984,950],{"class":940},[650,2986,2987],{"class":660},"\"anthropic/claude-sonnet-4-5-20250929\"",[650,2989,1991],{"class":940},[650,2991,2992,2995,2997,2999,3001,3004],{"class":652,"line":992},[650,2993,2994],{"class":664},"      \"researcher\"",[650,2996,1953],{"class":940},[650,2998,2982],{"class":664},[650,3000,950],{"class":940},[650,3002,3003],{"class":660},"\"openrouter/google/gemini-2.0-flash-001\"",[650,3005,3006],{"class":940}," }\n",[650,3008,3009],{"class":652,"line":1000},[650,3010,2033],{"class":940},[650,3012,3013],{"class":652,"line":1011},[650,3014,2038],{"class":940},[650,3016,3017],{"class":652,"line":1019},[650,3018,2043],{"class":940},[12,3020,3021],{},"This way, your default agent uses a free local model, but specialized agents like the coder can use a more capable cloud model when the task demands it.",[44,3023,3025],{"id":3024},"cost-comparison","Cost Comparison",[12,3027,3028],{},"Here's what typical monthly costs look like for moderate usage (~50 queries/day) across different setups:",[795,3030,3031,3046],{},[798,3032,3033],{},[801,3034,3035,3037,3040,3043],{},[804,3036,2299],{},[804,3038,3039],{},"AI Model Cost",[804,3041,3042],{},"Infrastructure",[804,3044,3045],{},"Total Monthly",[817,3047,3048,3062,3075,3089],{},[801,3049,3050,3053,3056,3059],{},[822,3051,3052],{},"Pure cloud (GPT / Gemini Pro)",[822,3054,3055],{},"$60–120",[822,3057,3058],{},"$5–10 VPS",[822,3060,3061],{},"$65–130",[801,3063,3064,3067,3070,3072],{},[822,3065,3066],{},"OpenRouter (Gemini Flash / Llama)",[822,3068,3069],{},"$10–30",[822,3071,3058],{},[822,3073,3074],{},"$15–40",[801,3076,3077,3080,3083,3086],{},[822,3078,3079],{},"Pure local (Ollama 32B+)",[822,3081,3082],{},"$0",[822,3084,3085],{},"$5–10 VPS + electricity",[822,3087,3088],{},"$5–15",[801,3090,3091,3094,3097,3099],{},[822,3092,3093],{},"Hybrid (local default + cloud for complex)",[822,3095,3096],{},"$10–25",[822,3098,3058],{},[822,3100,3101],{},"$15–35",[12,3103,3104,3105,3108],{},"The difference is dramatic. A hybrid setup can cut your monthly AI spend by ",[28,3106,3107],{},"70–85%"," compared to using cloud APIs exclusively.",[677,3110,3112],{"id":3111},"the-trade-offs","The Trade-offs",[12,3114,3115],{},"Local models aren't free of cost — they trade money for other things:",[112,3117,3118,3124,3130,3136,3142],{},[115,3119,3120,3123],{},[28,3121,3122],{},"Hardware requirements"," — you need 32 GB+ RAM or a decent GPU for OpenClaw-compatible models",[115,3125,3126,3129],{},[28,3127,3128],{},"Slower responses"," compared to cloud APIs, especially on CPU-only hardware",[115,3131,3132,3135],{},[28,3133,3134],{},"Lower quality"," on complex reasoning tasks compared to frontier cloud models",[115,3137,3138,3141],{},[28,3139,3140],{},"More setup work"," and occasional troubleshooting",[115,3143,3144,3147],{},[28,3145,3146],{},"Power consumption"," if running a GPU 24/7",[12,3149,3150],{},"For many users, the hybrid approach hits the sweet spot: fast and cheap for routine work, high quality when it matters.",[44,3152,3154],{"id":3153},"getting-started","Getting Started",[1547,3156,3157,3163,3174,3186,3192],{},[115,3158,3159,3162],{},[28,3160,3161],{},"Check your hardware"," — you need 32 GB+ RAM or a GPU with 24 GB+ VRAM for reliable results",[115,3164,3165,3167,3168,1720,3171],{},[28,3166,2312],{}," and pull a 32B+ model: ",[231,3169,3170],{},"ollama pull qwen3:32b",[231,3172,3173],{},"ollama pull deepseek-r1:32b",[115,3175,3176,3179,3180,3183,3184],{},[28,3177,3178],{},"Update your config"," — add the provider with ",[231,3181,3182],{},"\"api\": \"openai-completions\""," and a placeholder ",[231,3185,2072],{},[115,3187,3188,3191],{},[28,3189,3190],{},"Evaluate quality"," — run your typical tasks and see if the output is good enough",[115,3193,3194,3197],{},[28,3195,3196],{},"Go hybrid"," — configure per-agent models once you know which tasks need cloud quality",[12,3199,3200,3201,3206],{},"The OpenClaw community on ",[1584,3202,3205],{"href":3203,"rel":3204},"https://discord.gg/wTBAcqC38R",[1588],"Discord"," is a great place to share configs and get recommendations for which models work best for specific tasks.",[1591,3208],{},[12,3210,3211,3212,3216,3217,3220],{},"Want the convenience of managed hosting without the server management? ",[1584,3213,295],{"href":3214,"rel":3215},"https://dash.clawnest.ai/register",[1588]," handles infrastructure, updates, and backups so you can focus on configuring your AI assistants. Start with a ",[1584,3218,3219],{"href":1605},"free 3-day trial"," — no credit card required.",[1609,3222,3223],{},"html pre.shiki code .sScJk, html code.shiki .sScJk{--shiki-default:#6F42C1;--shiki-dark:#B392F0}html pre.shiki code .sj4cs, html code.shiki .sj4cs{--shiki-default:#005CC5;--shiki-dark:#79B8FF}html pre.shiki code .sZZnC, html code.shiki .sZZnC{--shiki-default:#032F62;--shiki-dark:#9ECBFF}html pre.shiki code .szBVR, html code.shiki .szBVR{--shiki-default:#D73A49;--shiki-dark:#F97583}html .default .shiki span {color: var(--shiki-default);background: var(--shiki-default-bg);font-style: var(--shiki-default-font-style);font-weight: var(--shiki-default-font-weight);text-decoration: var(--shiki-default-text-decoration);}html .shiki span {color: var(--shiki-default);background: var(--shiki-default-bg);font-style: var(--shiki-default-font-style);font-weight: var(--shiki-default-font-weight);text-decoration: var(--shiki-default-text-decoration);}html .dark .shiki span {color: var(--shiki-dark);background: var(--shiki-dark-bg);font-style: var(--shiki-dark-font-style);font-weight: var(--shiki-dark-font-weight);text-decoration: var(--shiki-dark-text-decoration);}html.dark .shiki span {color: var(--shiki-dark);background: var(--shiki-dark-bg);font-style: var(--shiki-dark-font-style);font-weight: var(--shiki-dark-font-weight);text-decoration: var(--shiki-dark-text-decoration);}html pre.shiki code .sVt8B, html code.shiki .sVt8B{--shiki-default:#24292E;--shiki-dark:#E1E4E8}",{"title":489,"searchDepth":490,"depth":490,"links":3225},[3226,3227,3228,3233,3236,3239,3242,3245,3248],{"id":1650,"depth":490,"text":1651},{"id":1679,"depth":490,"text":1680},{"id":1739,"depth":490,"text":1740,"children":3229},[3230,3231,3232],{"id":1751,"depth":944,"text":1752},{"id":1800,"depth":944,"text":1801},{"id":2110,"depth":944,"text":2111},{"id":2168,"depth":490,"text":2169,"children":3234},[3235],{"id":2252,"depth":944,"text":2253},{"id":2286,"depth":490,"text":2287,"children":3237},[3238],{"id":2298,"depth":944,"text":2299},{"id":2550,"depth":490,"text":2551,"children":3240},[3241],{"id":2575,"depth":944,"text":2576},{"id":2901,"depth":490,"text":2902,"children":3243},[3244],{"id":2908,"depth":944,"text":2909},{"id":3024,"depth":490,"text":3025,"children":3246},[3247],{"id":3111,"depth":944,"text":3112},{"id":3153,"depth":490,"text":3154},"2026-02-15","A practical guide to running local LLMs with OpenClaw using Ollama. Covers setup, config, hardware requirements, OpenRouter, and hybrid strategies to reduce your monthly AI spend.","/blog/ollama-clawnest-openclaw.png",{},"/blog/openclaw-ollama-local-models-guide",{"title":1645,"description":3250},"blog/openclaw-ollama-local-models-guide",[1747,3257,3258,3259],"Cost Optimization","Self-Hosting","Guide","dLsc1xXYd6PA5ng5P-sH-yUtLyVpDeKVKrcoD-4J_dU",{"id":3262,"title":3263,"author":7,"body":3264,"date":4084,"description":4085,"draft":506,"extension":507,"image":4086,"meta":4087,"navigation":510,"path":4088,"seo":4089,"stem":4090,"tags":4091,"__hash__":4094},"blog/blog/openclaw-vs-self-hosting.md","OpenClaw: Managed Hosting vs Self-Hosting — What You Should Know",{"type":9,"value":3265,"toc":4060},[3266,3270,3273,3279,3282,3286,3289,3291,3298,3369,3372,3376,3379,3433,3440,3444,3447,3497,3501,3508,3529,3532,3536,3539,3543,3656,3659,3674,3681,3683,3686,3690,3748,3752,3798,3801,3805,3808,3866,3869,3873,3880,3884,3887,3925,3929,3979,3983,3986,4003,4007,4010,4030,4034,4037,4040,4047,4049],[44,3267,3269],{"id":3268},"the-hosting-question","The Hosting Question",[12,3271,3272],{},"OpenClaw is open source — you can run it anywhere. That's one of its strengths. But \"can run anywhere\" doesn't mean every option is equal in terms of effort, cost, and reliability.",[12,3274,3275,3276],{},"Since OpenClaw went from 9,000 to 60,000+ GitHub stars in January 2026, thousands of new users face the same question: ",[28,3277,3278],{},"should I self-host or use managed hosting?",[12,3280,3281],{},"This guide compares both approaches honestly, including where self-hosting wins.",[44,3283,3285],{"id":3284},"what-self-hosting-actually-involves","What Self-Hosting Actually Involves",[12,3287,3288],{},"Self-hosting means you rent a VPS (or use local hardware), install OpenClaw yourself, and take responsibility for everything that keeps it running.",[677,3290,1143],{"id":1142},[12,3292,3293,3294,3297],{},"The OpenClaw docs specify these ",[28,3295,3296],{},"minimum requirements"," for a VPS:",[795,3299,3300,3313],{},[798,3301,3302],{},[801,3303,3304,3307,3310],{},[804,3305,3306],{},"Resource",[804,3308,3309],{},"Minimum",[804,3311,3312],{},"Recommended",[817,3314,3315,3326,3337,3348,3359],{},[801,3316,3317,3320,3323],{},[822,3318,3319],{},"CPU",[822,3321,3322],{},"1 vCPU",[822,3324,3325],{},"2+ vCPU",[801,3327,3328,3331,3334],{},[822,3329,3330],{},"RAM",[822,3332,3333],{},"2 GB",[822,3335,3336],{},"4 GB+",[801,3338,3339,3342,3345],{},[822,3340,3341],{},"Storage",[822,3343,3344],{},"20 GB SSD",[822,3346,3347],{},"40 GB NVMe",[801,3349,3350,3353,3356],{},[822,3351,3352],{},"OS",[822,3354,3355],{},"Ubuntu 22.04+",[822,3357,3358],{},"Ubuntu 24.04",[801,3360,3361,3364,3367],{},[822,3362,3363],{},"Runtime",[822,3365,3366],{},"Node.js 22+",[822,3368,3366],{},[12,3370,3371],{},"Docker is the recommended deployment method, which adds its own resource overhead.",[677,3373,3375],{"id":3374},"the-setup-process","The Setup Process",[12,3377,3378],{},"A typical self-hosted OpenClaw deployment looks like this:",[1547,3380,3381,3387,3393,3399,3405,3411,3421,3427],{},[115,3382,3383,3386],{},[28,3384,3385],{},"Provision a VPS"," — choose a provider (Hetzner, DigitalOcean, AWS, etc.), select a plan, and set up SSH access",[115,3388,3389,3392],{},[28,3390,3391],{},"Secure the server"," — configure firewall rules, disable password authentication, set up Fail2Ban, enable automatic security updates",[115,3394,3395,3398],{},[28,3396,3397],{},"Install Docker"," — install Docker Engine and Docker Compose",[115,3400,3401,3404],{},[28,3402,3403],{},"Install OpenClaw"," — pull the image or install via npm, configure the gateway",[115,3406,3407,3410],{},[28,3408,3409],{},"Set up networking"," — install and configure Nginx as a reverse proxy, obtain SSL certificates via Let's Encrypt, configure DNS",[115,3412,3413,3416,3417,3420],{},[28,3414,3415],{},"Configure OpenClaw"," — edit ",[231,3418,3419],{},"~/.openclaw/openclaw.json",", set up channels, install skills",[115,3422,3423,3426],{},[28,3424,3425],{},"Set up backups"," — write backup scripts, configure cron jobs, set up off-site backup storage",[115,3428,3429,3432],{},[28,3430,3431],{},"Set up monitoring"," — configure uptime checks, log rotation, alerting",[12,3434,3435,3436,3439],{},"For experienced Linux administrators, this takes ",[28,3437,3438],{},"2–6 hours",". For users learning as they go, it can take a full weekend or more.",[677,3441,3443],{"id":3442},"ongoing-maintenance","Ongoing Maintenance",[12,3445,3446],{},"The initial setup is just the beginning. Self-hosting means ongoing responsibility for:",[112,3448,3449,3455,3461,3467,3473,3479,3485,3491],{},[115,3450,3451,3454],{},[28,3452,3453],{},"OS updates"," — security patches, kernel updates, reboots",[115,3456,3457,3460],{},[28,3458,3459],{},"Docker updates"," — keeping Docker and images current",[115,3462,3463,3466],{},[28,3464,3465],{},"OpenClaw updates"," — manually pulling new versions, testing, and deploying",[115,3468,3469,3472],{},[28,3470,3471],{},"SSL renewal"," — Let's Encrypt certificates expire every 90 days (auto-renewal needs configuration)",[115,3474,3475,3478],{},[28,3476,3477],{},"Backup verification"," — regularly testing that backups actually work",[115,3480,3481,3484],{},[28,3482,3483],{},"Security monitoring"," — watching for unauthorized access attempts, keeping Fail2Ban rules current",[115,3486,3487,3490],{},[28,3488,3489],{},"Disk management"," — logs and Docker images fill up disk space over time",[115,3492,3493,3496],{},[28,3494,3495],{},"Troubleshooting"," — when something breaks at 2 AM, it's your problem",[677,3498,3500],{"id":3499},"security-considerations","Security Considerations",[12,3502,3503,3504,3507],{},"The OpenClaw documentation explicitly warns: because OpenClaw has ",[28,3505,3506],{},"direct access to your terminal and file system",", deploying it in a strictly isolated environment is critical. The recommended security measures include:",[112,3509,3510,3517,3520,3523,3526],{},[115,3511,3512,3513,3516],{},"Running on a ",[28,3514,3515],{},"dedicated machine or VPS"," with no sensitive personal data",[115,3518,3519],{},"Blocking direct access to OpenClaw's port 3000, using Nginx to handle external traffic",[115,3521,3522],{},"SSH key authentication only, with password login disabled",[115,3524,3525],{},"Fail2Ban for brute-force protection",[115,3527,3528],{},"Automatic security updates enabled",[12,3530,3531],{},"Misconfiguring any of these exposes your server — and potentially your API keys and personal data — to the internet.",[44,3533,3535],{"id":3534},"what-managed-hosting-provides","What Managed Hosting Provides",[12,3537,3538],{},"With managed hosting, you get a private OpenClaw instance that's always on, always updated, and accessible from any device. The provider handles infrastructure, security, and maintenance.",[677,3540,3542],{"id":3541},"what-clawnest-manages-for-you","What ClawNest Manages for You",[795,3544,3545,3557],{},[798,3546,3547],{},[801,3548,3549,3552,3555],{},[804,3550,3551],{},"Responsibility",[804,3553,3554],{},"Self-Hosted",[804,3556,295],{},[817,3558,3559,3570,3580,3590,3599,3608,3616,3627,3636,3647],{},[801,3560,3561,3564,3567],{},[822,3562,3563],{},"Server provisioning",[822,3565,3566],{},"You",[822,3568,3569],{},"Managed",[801,3571,3572,3575,3577],{},[822,3573,3574],{},"OS updates & patching",[822,3576,3566],{},[822,3578,3579],{},"Automatic",[801,3581,3582,3585,3587],{},[822,3583,3584],{},"Docker setup",[822,3586,3566],{},[822,3588,3589],{},"Not needed",[801,3591,3592,3595,3597],{},[822,3593,3594],{},"SSL certificates",[822,3596,3566],{},[822,3598,3579],{},[801,3600,3601,3604,3606],{},[822,3602,3603],{},"Firewall & security",[822,3605,3566],{},[822,3607,3569],{},[801,3609,3610,3612,3614],{},[822,3611,3465],{},[822,3613,3566],{},[822,3615,3579],{},[801,3617,3618,3621,3624],{},[822,3619,3620],{},"Backups",[822,3622,3623],{},"You (scripts + storage)",[822,3625,3626],{},"Automatic daily",[801,3628,3629,3632,3634],{},[822,3630,3631],{},"Monitoring & alerting",[822,3633,3566],{},[822,3635,1410],{},[801,3637,3638,3641,3644],{},[822,3639,3640],{},"Uptime",[822,3642,3643],{},"Best-effort",[822,3645,3646],{},"24/7 infrastructure",[801,3648,3649,3652,3654],{},[822,3650,3651],{},"Disaster recovery",[822,3653,3566],{},[822,3655,3569],{},[677,3657,3375],{"id":3658},"the-setup-process-1",[1547,3660,3661,3668,3671],{},[115,3662,3663,3664],{},"Sign up at ",[1584,3665,3667],{"href":3214,"rel":3666},[1588],"dash.clawnest.ai",[115,3669,3670],{},"Create an AI assistant, paste your API key, configure personality",[115,3672,3673],{},"Click Deploy",[12,3675,3676,3677,3680],{},"Time: ",[28,3678,3679],{},"under 5 minutes",". No terminal, no Docker, no SSH.",[44,3682,3025],{"id":3024},[12,3684,3685],{},"Cost is where self-hosting looks most attractive on paper. Let's break down the real numbers.",[677,3687,3689],{"id":3688},"self-hosted-vps","Self-Hosted VPS",[795,3691,3692,3702],{},[798,3693,3694],{},[801,3695,3696,3699],{},[804,3697,3698],{},"Item",[804,3700,3701],{},"Monthly Cost",[817,3703,3704,3712,3720,3728,3740],{},[801,3705,3706,3709],{},[822,3707,3708],{},"VPS (2 GB RAM, Hetzner)",[822,3710,3711],{},"~€4–8",[801,3713,3714,3717],{},[822,3715,3716],{},"Domain name",[822,3718,3719],{},"~€1 (amortized)",[801,3721,3722,3725],{},[822,3723,3724],{},"Backup storage (optional)",[822,3726,3727],{},"€1–5",[801,3729,3730,3735],{},[822,3731,3732],{},[28,3733,3734],{},"Infrastructure total",[822,3736,3737],{},[28,3738,3739],{},"€6–14",[801,3741,3742,3745],{},[822,3743,3744],{},"Your time (setup + maintenance)",[822,3746,3747],{},"2–4 hours/month",[677,3749,3751],{"id":3750},"clawnest-starter-plan","ClawNest Starter Plan",[795,3753,3754,3762],{},[798,3755,3756],{},[801,3757,3758,3760],{},[804,3759,3698],{},[804,3761,3701],{},[817,3763,3764,3772,3780,3790],{},[801,3765,3766,3769],{},[822,3767,3768],{},"Managed hosting (1 AI assistant)",[822,3770,3771],{},"€12",[801,3773,3774,3777],{},[822,3775,3776],{},"Backups, updates, monitoring",[822,3778,3779],{},"Included",[801,3781,3782,3786],{},[822,3783,3784],{},[28,3785,3734],{},[822,3787,3788],{},[28,3789,3771],{},[801,3791,3792,3795],{},[822,3793,3794],{},"Your time",[822,3796,3797],{},"~0 hours/month",[12,3799,3800],{},"The raw infrastructure cost of self-hosting can be lower — by roughly €4–8 per month. But this ignores the value of your time. If you spend 3 hours per month on maintenance and your time is worth even €15/hour, self-hosting actually costs €51–59/month in total.",[677,3802,3804],{"id":3803},"for-multiple-ai-assistants","For Multiple AI Assistants",[12,3806,3807],{},"On a VPS, you can run multiple OpenClaw instances on the same server (as long as resources allow). On ClawNest, additional AI assistants start from €8/month each.",[795,3809,3810,3823],{},[798,3811,3812],{},[801,3813,3814,3817,3820],{},[804,3815,3816],{},"AI Assistants",[804,3818,3819],{},"VPS Cost",[804,3821,3822],{},"ClawNest Cost",[817,3824,3825,3834,3844,3855],{},[801,3826,3827,3830,3832],{},[822,3828,3829],{},"1",[822,3831,3739],{},[822,3833,3771],{},[801,3835,3836,3839,3841],{},[822,3837,3838],{},"2",[822,3840,3739],{},[822,3842,3843],{},"€21",[801,3845,3846,3849,3852],{},[822,3847,3848],{},"3",[822,3850,3851],{},"€10–20 (may need bigger VPS)",[822,3853,3854],{},"€30",[801,3856,3857,3860,3863],{},[822,3858,3859],{},"5",[822,3861,3862],{},"€15–30",[822,3864,3865],{},"€48",[12,3867,3868],{},"For users running 5+ AI assistants, self-hosting becomes more cost-effective in raw infrastructure terms — if you have the skills and time.",[677,3870,3872],{"id":3871},"ai-token-costs","AI Token Costs",[12,3874,3875,3876,3879],{},"Both approaches use ",[28,3877,3878],{},"BYOK (Bring Your Own Key)",", so AI token costs are identical. You pay Anthropic or OpenAI directly at their published rates. ClawNest adds zero markup on AI usage.",[44,3881,3883],{"id":3882},"where-self-hosting-wins","Where Self-Hosting Wins",[12,3885,3886],{},"Let's be honest about the advantages:",[112,3888,3889,3895,3901,3907,3913,3919],{},[115,3890,3891,3894],{},[28,3892,3893],{},"Full root access"," — complete control over the server, OS, and network configuration",[115,3896,3897,3900],{},[28,3898,3899],{},"Custom networking"," — Tailscale, VPN tunnels, custom DNS, and advanced routing",[115,3902,3903,3906],{},[28,3904,3905],{},"Local AI models"," — run Ollama or similar for on-device inference (requires 16 GB+ RAM)",[115,3908,3909,3912],{},[28,3910,3911],{},"No vendor dependency"," — you own and control everything",[115,3914,3915,3918],{},[28,3916,3917],{},"Cost at scale"," — running many AI assistants on one server can be cheaper per-assistant",[115,3920,3921,3924],{},[28,3922,3923],{},"Learning opportunity"," — great way to build Linux and DevOps skills",[44,3926,3928],{"id":3927},"where-managed-hosting-wins","Where Managed Hosting Wins",[112,3930,3931,3937,3943,3949,3955,3961,3967,3973],{},[115,3932,3933,3936],{},[28,3934,3935],{},"Zero setup time"," — running in minutes, not hours",[115,3938,3939,3942],{},[28,3940,3941],{},"Zero maintenance"," — no patching, no Docker, no certificate management",[115,3944,3945,3948],{},[28,3946,3947],{},"Automatic updates"," — new OpenClaw versions deployed without downtime",[115,3950,3951,3954],{},[28,3952,3953],{},"Automatic backups"," — daily backups with one-click restore",[115,3956,3957,3960],{},[28,3958,3959],{},"Professional security"," — hardened infrastructure without configuring it yourself",[115,3962,3963,3966],{},[28,3964,3965],{},"EU data residency"," — hosted in Sweden, GDPR-aligned by default",[115,3968,3969,3972],{},[28,3970,3971],{},"Dashboard + GUI"," — manage everything visually",[115,3974,3975,3978],{},[28,3976,3977],{},"Predictable cost"," — same bill every month, no surprise resource usage",[44,3980,3982],{"id":3981},"who-should-self-host","Who Should Self-Host?",[12,3984,3985],{},"Self-hosting makes sense if you:",[112,3987,3988,3991,3994,3997,4000],{},[115,3989,3990],{},"Are an experienced Linux administrator who enjoys server management",[115,3992,3993],{},"Need full root access for custom networking or local AI models",[115,3995,3996],{},"Run 5+ instances and want to minimize infrastructure cost",[115,3998,3999],{},"Want to learn DevOps and don't mind investing time",[115,4001,4002],{},"Have compliance requirements that mandate your own hardware",[44,4004,4006],{"id":4005},"who-should-use-managed-hosting","Who Should Use Managed Hosting?",[12,4008,4009],{},"Managed hosting makes sense if you:",[112,4011,4012,4015,4018,4021,4024,4027],{},[115,4013,4014],{},"Want OpenClaw running 24/7 without managing servers",[115,4016,4017],{},"Value your time over saving a few euros per month",[115,4019,4020],{},"Don't have Linux/Docker experience (or don't want to use it for this)",[115,4022,4023],{},"Need reliable backups and automatic updates",[115,4025,4026],{},"Want EU data residency without configuring it yourself",[115,4028,4029],{},"Are running a business and need predictable costs",[44,4031,4033],{"id":4032},"the-honest-take","The Honest Take",[12,4035,4036],{},"Self-hosting is a legitimate option — OpenClaw is open source specifically so you can run it however you want. If you're a Linux pro who enjoys infrastructure work, go for it.",[12,4038,4039],{},"But for most people, the math favors managed hosting. The €12/month for ClawNest's Starter plan buys you hours of time back every month, professional-grade security you'd have to configure yourself, and the peace of mind that your assistant is always running and always backed up.",[12,4041,4042,4043,4046],{},"The best part of BYOK hosting: ",[28,4044,4045],{},"your data stays yours either way",". ClawNest never accesses your AI conversations or API keys beyond what's needed to run the service.",[1591,4048],{},[12,4050,4051,4052,4054,4055,4059],{},"Ready to skip the server setup? Start with a ",[1584,4053,3219],{"href":1605}," — no credit card required. Already self-hosting and want to migrate? ",[1584,4056,4058],{"href":3203,"rel":4057},[1588],"Reach out on Discord"," and we'll help you move over.",{"title":489,"searchDepth":490,"depth":490,"links":4061},[4062,4063,4069,4073,4079,4080,4081,4082,4083],{"id":3268,"depth":490,"text":3269},{"id":3284,"depth":490,"text":3285,"children":4064},[4065,4066,4067,4068],{"id":1142,"depth":944,"text":1143},{"id":3374,"depth":944,"text":3375},{"id":3442,"depth":944,"text":3443},{"id":3499,"depth":944,"text":3500},{"id":3534,"depth":490,"text":3535,"children":4070},[4071,4072],{"id":3541,"depth":944,"text":3542},{"id":3658,"depth":944,"text":3375},{"id":3024,"depth":490,"text":3025,"children":4074},[4075,4076,4077,4078],{"id":3688,"depth":944,"text":3689},{"id":3750,"depth":944,"text":3751},{"id":3803,"depth":944,"text":3804},{"id":3871,"depth":944,"text":3872},{"id":3882,"depth":490,"text":3883},{"id":3927,"depth":490,"text":3928},{"id":3981,"depth":490,"text":3982},{"id":4005,"depth":490,"text":4006},{"id":4032,"depth":490,"text":4033},"2026-02-14","A honest comparison of self-hosting OpenClaw on a VPS versus using managed hosting. Covers setup, costs, security, maintenance, and who each option is best for.","/blog/managed-hosting-clawnest.png",{},"/blog/openclaw-vs-self-hosting",{"title":3263,"description":4085},"blog/openclaw-vs-self-hosting",[4092,3258,4093],"Hosting","Comparison","Vyf29CNujtQm0ckS2PlOa4mweBm05Xgv7sV7c9-Al-Q",{"id":4096,"title":4097,"author":7,"body":4098,"date":4647,"description":4648,"draft":506,"extension":507,"image":4649,"meta":4650,"navigation":510,"path":4651,"seo":4652,"stem":4653,"tags":4654,"__hash__":4657},"blog/blog/deploy-openclaw-clawnest.md","How to Deploy OpenClaw on ClawNest in 5 Minutes",{"type":9,"value":4099,"toc":4620},[4100,4104,4107,4110,4113,4117,4131,4134,4138,4141,4146,4165,4170,4185,4191,4195,4206,4209,4213,4216,4233,4236,4240,4243,4247,4250,4271,4275,4278,4310,4317,4321,4324,4338,4341,4345,4348,4379,4386,4390,4396,4422,4425,4429,4432,4502,4513,4517,4520,4524,4527,4531,4534,4560,4564,4567,4571,4574,4577,4581,4584,4588,4591,4595,4606,4608],[44,4101,4103],{"id":4102},"why-managed-hosting","Why Managed Hosting?",[12,4105,4106],{},"Self-hosting OpenClaw means setting up a VPS, installing Docker, configuring Node.js 22+, managing SSL certificates, setting up reverse proxies, and keeping everything patched and running. It's a multi-hour project for experienced Linux users — and an ongoing maintenance commitment.",[12,4108,4109],{},"With ClawNest, you skip all of that. Your OpenClaw instance runs on managed infrastructure in the EU (Sweden), with automatic backups, monitoring, and updates handled for you.",[12,4111,4112],{},"Here's how to go from zero to a running assistant in about 5 minutes.",[44,4114,4116],{"id":4115},"what-youll-need","What You'll Need",[112,4118,4119,4125],{},[115,4120,4121,4124],{},[28,4122,4123],{},"An email address"," — for your ClawNest account",[115,4126,4127,4130],{},[28,4128,4129],{},"An API key"," from Anthropic or OpenAI",[12,4132,4133],{},"That's it. No server, no Docker, no SSH keys, no domain name.",[677,4135,4137],{"id":4136},"getting-an-api-key","Getting an API Key",[12,4139,4140],{},"If you don't have an API key yet:",[12,4142,4143],{},[28,4144,4145],{},"Anthropic (Claude) — recommended:",[1547,4147,4148,4156,4159,4162],{},[115,4149,4150,4151],{},"Go to ",[1584,4152,4155],{"href":4153,"rel":4154},"https://console.anthropic.com",[1588],"console.anthropic.com",[115,4157,4158],{},"Create an account and add billing",[115,4160,4161],{},"Navigate to API Keys and create a new key",[115,4163,4164],{},"Copy it somewhere safe",[12,4166,4167],{},[28,4168,4169],{},"OpenAI:",[1547,4171,4172,4179,4181,4183],{},[115,4173,4150,4174],{},[1584,4175,4178],{"href":4176,"rel":4177},"https://platform.openai.com",[1588],"platform.openai.com",[115,4180,4158],{},[115,4182,4161],{},[115,4184,4164],{},[12,4186,4187,4188,4190],{},"ClawNest uses ",[28,4189,3878],{}," — you pay providers directly at their published rates. We add zero markup on AI tokens.",[44,4192,4194],{"id":4193},"step-1-create-your-clawnest-account","Step 1: Create Your ClawNest Account",[12,4196,4197,4198,4201,4202,4205],{},"Head to ",[1584,4199,3667],{"href":3214,"rel":4200},[1588]," and sign up. The free trial gives you ",[28,4203,4204],{},"3 days of full access"," with one AI assistant instance — no credit card required.",[12,4207,4208],{},"You'll get access to the full dashboard immediately after email verification.",[44,4210,4212],{"id":4211},"step-2-create-a-new-ai-assistant-instance","Step 2: Create a New AI Assistant Instance",[12,4214,4215],{},"From the dashboard:",[1547,4217,4218,4224,4227,4230],{},[115,4219,4220,4221],{},"Click ",[28,4222,4223],{},"\"New AI Assistant\"",[115,4225,4226],{},"Give your instance a name (e.g., \"Personal Assistant\" or \"Support AI Assistant\")",[115,4228,4229],{},"Paste your API key",[115,4231,4232],{},"Select your preferred AI model",[12,4234,4235],{},"The dashboard validates your API key in real-time. If there's an issue with your key, you'll know immediately.",[44,4237,4239],{"id":4238},"step-3-configure-your-assistant","Step 3: Configure Your Assistant",[12,4241,4242],{},"Use the visual editor to customize your assistant's behavior:",[677,4244,4246],{"id":4245},"personality","Personality",[12,4248,4249],{},"Define how your assistant should behave. This is the system prompt that shapes every response. Examples:",[112,4251,4252,4257,4266],{},[115,4253,4254],{},[1596,4255,4256],{},"\"You are a helpful personal assistant. Be concise and proactive.\"",[115,4258,4259],{},[1596,4260,4261,4262,4265],{},"\"You are a customer support agent for ",[650,4263,4264],{},"company",". Always be professional and refer to the knowledge base.\"",[115,4267,4268],{},[1596,4269,4270],{},"\"You are a research analyst. Provide sourced, detailed answers.\"",[677,4272,4274],{"id":4273},"skills","Skills",[12,4276,4277],{},"Enable the capabilities your assistant needs:",[112,4279,4280,4286,4292,4298,4304],{},[115,4281,4282,4285],{},[28,4283,4284],{},"Web search"," — find information online",[115,4287,4288,4291],{},[28,4289,4290],{},"File handling"," — read, write, and manage files",[115,4293,4294,4297],{},[28,4295,4296],{},"Code execution"," — run code and scripts",[115,4299,4300,4303],{},[28,4301,4302],{},"Calendar integration"," — manage schedules",[115,4305,4306,4309],{},[28,4307,4308],{},"Email tools"," — draft and manage email",[12,4311,4312,4313,4316],{},"ClawNest provides access to curated skills from ",[28,4314,4315],{},"ClawHub",", the OpenClaw skill registry with 5,700+ community-built extensions. Install any skill with a single click.",[677,4318,4320],{"id":4319},"rules-and-guardrails","Rules and Guardrails",[12,4322,4323],{},"Set boundaries for what your assistant should and shouldn't do:",[112,4325,4326,4329,4332,4335],{},[115,4327,4328],{},"Topics to avoid",[115,4330,4331],{},"Response length limits",[115,4333,4334],{},"Escalation rules (when to hand off to a human)",[115,4336,4337],{},"Data handling policies",[12,4339,4340],{},"No coding required — everything is configured through the dashboard UI.",[44,4342,4344],{"id":4343},"step-4-connect-channels-optional","Step 4: Connect Channels (Optional)",[12,4346,4347],{},"If you want to reach your assistant through messaging platforms, connect one or more channels:",[112,4349,4350,4356,4362,4367,4373],{},[115,4351,4352,4355],{},[28,4353,4354],{},"WhatsApp"," — personal and business messaging",[115,4357,4358,4361],{},[28,4359,4360],{},"Telegram"," — fast, lightweight chat",[115,4363,4364,4366],{},[28,4365,3205],{}," — community and team channels",[115,4368,4369,4372],{},[28,4370,4371],{},"Slack"," — workplace integration",[115,4374,4375,4378],{},[28,4376,4377],{},"Microsoft Teams"," — enterprise environments",[12,4380,4381,4382,4385],{},"Each channel has a guided setup flow in the dashboard. You can also skip this step and use the built-in ",[28,4383,4384],{},"WebChat"," interface to start chatting immediately.",[44,4387,4389],{"id":4388},"step-5-deploy","Step 5: Deploy",[12,4391,4220,4392,4395],{},[28,4393,4394],{},"\"Deploy\""," and your assistant goes live. It's now running 24/7 on ClawNest infrastructure with:",[112,4397,4398,4404,4410,4416],{},[115,4399,4400,4403],{},[28,4401,4402],{},"Automatic daily backups"," with one-click restore",[115,4405,4406,4409],{},[28,4407,4408],{},"GUI access"," for configuration changes",[115,4411,4412,4415],{},[28,4413,4414],{},"Status monitoring"," from the dashboard",[115,4417,4418,4421],{},[28,4419,4420],{},"Auto-updates"," when new OpenClaw versions are released",[12,4423,4424],{},"Deployment typically takes under 30 seconds.",[44,4426,4428],{"id":4427},"what-your-12month-gets-you","What Your €12/Month Gets You",[12,4430,4431],{},"On the Starter plan, you get:",[795,4433,4434,4442],{},[798,4435,4436],{},[801,4437,4438,4440],{},[804,4439,1250],{},[804,4441,3779],{},[817,4443,4444,4451,4458,4464,4472,4478,4486,4494],{},[801,4445,4446,4449],{},[822,4447,4448],{},"AI assistant instances",[822,4450,3829],{},[801,4452,4453,4455],{},[822,4454,3640],{},[822,4456,4457],{},"24/7",[801,4459,4460,4462],{},[822,4461,3620],{},[822,4463,3626],{},[801,4465,4466,4469],{},[822,4467,4468],{},"Dashboard access",[822,4470,4471],{},"Full",[801,4473,4474,4476],{},[822,4475,3465],{},[822,4477,3579],{},[801,4479,4480,4483],{},[822,4481,4482],{},"Data location",[822,4484,4485],{},"EU (Sweden)",[801,4487,4488,4491],{},[822,4489,4490],{},"Support",[822,4492,4493],{},"Discord community",[801,4495,4496,4499],{},[822,4497,4498],{},"Additional AI assistants",[822,4500,4501],{},"From €8/mo each",[12,4503,4504,4505,4508,4509,4512],{},"Need more? The ",[1584,4506,4507],{"href":1605},"Pro plan (€25/mo)"," adds shell access, priority support, and advanced monitoring. The ",[1584,4510,4511],{"href":1605},"Team plan (€99/mo)"," includes multiple AI assistant instances and team management.",[44,4514,4516],{"id":4515},"after-deployment","After Deployment",[12,4518,4519],{},"Once your assistant is running, here are some next steps:",[677,4521,4523],{"id":4522},"fine-tune-the-personality","Fine-Tune the Personality",[12,4525,4526],{},"Talk to your assistant and iterate on the system prompt. Most users refine the personality over the first few days as they discover what works best for their use case.",[677,4528,4530],{"id":4529},"add-more-skills","Add More Skills",[12,4532,4533],{},"Browse ClawHub from the dashboard and enable skills as you need them. Popular picks:",[112,4535,4536,4542,4548,4554],{},[115,4537,4538,4541],{},[231,4539,4540],{},"google-calendar"," — schedule management",[115,4543,4544,4547],{},[231,4545,4546],{},"web-search"," — real-time information lookup",[115,4549,4550,4553],{},[231,4551,4552],{},"github"," — repository management and code review",[115,4555,4556,4559],{},[231,4557,4558],{},"notion"," — knowledge base integration",[677,4561,4563],{"id":4562},"monitor-usage","Monitor Usage",[12,4565,4566],{},"Track your assistant's activity, response times, and API consumption from the dashboard. This helps you optimize model selection and control costs.",[677,4568,4570],{"id":4569},"scale-up","Scale Up",[12,4572,4573],{},"Need more AI assistants? Add them from €8/mo each as add-ons. Each additional AI assistant gets its own isolated instance with separate configuration and memory.",[44,4575,3495],{"id":4576},"troubleshooting",[677,4578,4580],{"id":4579},"api-key-not-working","API Key Not Working",[12,4582,4583],{},"Make sure your API key is active and has sufficient credits with your provider. You can test the connection directly from the ClawNest dashboard.",[677,4585,4587],{"id":4586},"ai-assistant-not-responding","AI Assistant Not Responding",[12,4589,4590],{},"Check the AI assistant status in your dashboard. If it shows as offline, try redeploying from the dashboard — it takes seconds. For persistent issues, check your API key balance.",[677,4592,4594],{"id":4593},"need-help","Need Help?",[12,4596,4597,4598,4601,4602,157],{},"Our primary support channel is ",[1584,4599,3205],{"href":3203,"rel":4600},[1588],". The community and team are active and happy to help with setup questions. You can also reach us at ",[1584,4603,4605],{"href":4604},"mailto:contact@clawnest.ai","contact@clawnest.ai",[1591,4607],{},[12,4609,4610,4611,4615,4616,157],{},"Ready to start? ",[1584,4612,4614],{"href":3214,"rel":4613},[1588],"Create your free account"," and have your OpenClaw assistant running in 5 minutes. For the detailed feature breakdown, visit our ",[1584,4617,4619],{"href":4618},"/features","features page",{"title":489,"searchDepth":490,"depth":490,"links":4621},[4622,4623,4626,4627,4628,4633,4634,4635,4636,4642],{"id":4102,"depth":490,"text":4103},{"id":4115,"depth":490,"text":4116,"children":4624},[4625],{"id":4136,"depth":944,"text":4137},{"id":4193,"depth":490,"text":4194},{"id":4211,"depth":490,"text":4212},{"id":4238,"depth":490,"text":4239,"children":4629},[4630,4631,4632],{"id":4245,"depth":944,"text":4246},{"id":4273,"depth":944,"text":4274},{"id":4319,"depth":944,"text":4320},{"id":4343,"depth":490,"text":4344},{"id":4388,"depth":490,"text":4389},{"id":4427,"depth":490,"text":4428},{"id":4515,"depth":490,"text":4516,"children":4637},[4638,4639,4640,4641],{"id":4522,"depth":944,"text":4523},{"id":4529,"depth":944,"text":4530},{"id":4562,"depth":944,"text":4563},{"id":4569,"depth":944,"text":4570},{"id":4576,"depth":490,"text":3495,"children":4643},[4644,4645,4646],{"id":4579,"depth":944,"text":4580},{"id":4586,"depth":944,"text":4587},{"id":4593,"depth":944,"text":4594},"2026-02-12","Step-by-step guide to deploying your own OpenClaw AI assistant on ClawNest managed hosting — from sign-up to a running instance in under 5 minutes.",null,{},"/blog/deploy-openclaw-clawnest",{"title":4097,"description":4648},"blog/deploy-openclaw-clawnest",[4655,4656,295],"Tutorial","Deployment","km1RP2YFAYiMen7xJjyS7dSu22cd4CACp2n3jxTmnOM",{"id":4659,"title":4660,"author":7,"body":4661,"date":5178,"description":5179,"draft":506,"extension":507,"image":4649,"meta":5180,"navigation":510,"path":5181,"seo":5182,"stem":5183,"tags":5184,"__hash__":5186},"blog/blog/what-is-openclaw.md","What Is OpenClaw? A Complete Guide to the Open-Source AI Assistant",{"type":9,"value":4662,"toc":5161},[4663,4667,4678,4685,4688,4692,4699,4705,4708,4710,4714,4717,4762,4765,4769,4772,4786,4789,4793,4803,4809,4826,4829,4848,4855,4859,4862,4876,4882,4886,4889,4893,4896,4900,4903,4967,4973,4977,4980,5051,5054,5069,5075,5079,5082,5092,5098,5102,5105,5137,5145,5147,5158],[44,4664,4666],{"id":4665},"what-is-openclaw","What Is OpenClaw?",[12,4668,4669,4670,4673,4674,4677],{},"OpenClaw is a free, open-source personal AI assistant that runs on your own devices. Created by Austrian developer Peter Steinberger and originally released in November 2025 under the name ",[1596,4671,4672],{},"Clawdbot",", it was briefly renamed ",[1596,4675,4676],{},"Moltbot"," before settling on OpenClaw in late January 2026.",[12,4679,4680,4681,4684],{},"Unlike cloud-only assistants like ChatGPT or Gemini, OpenClaw runs ",[28,4682,4683],{},"locally"," — your files, emails, and personal data never leave your machine unless you tell it to reach out. Because it has direct access to your operating system, it can browse the web, read and write files, run shell commands, and take autonomous actions on your behalf.",[12,4686,4687],{},"The project exploded in popularity in January 2026, jumping from 9,000 to over 60,000 GitHub stars in a matter of days, making it one of the fastest-growing open-source projects in history.",[44,4689,4691],{"id":4690},"how-openclaw-works","How OpenClaw Works",[12,4693,4694,4695,4698],{},"At its core, OpenClaw is a ",[28,4696,4697],{},"local gateway"," — a WebSocket-based control plane that coordinates sessions, messaging channels, tools, and events. It runs as a daemon on your machine and connects to the AI providers you choose.",[642,4700,4703],{"className":4701,"code":4702,"language":709},[707],"Your Channels (WhatsApp, Slack, Discord...)\n        ↓\n   OpenClaw Gateway (ws://127.0.0.1:18789)\n        ↓\n   LLM Provider (Anthropic, OpenAI)\n        ↓\n   Skills & Tools (files, browser, APIs...)\n",[231,4704,4702],{"__ignoreMap":489},[12,4706,4707],{},"You interact with your assistant through the messaging platforms you already use. OpenClaw handles the orchestration between your messages, the AI model, and whatever tools or skills are needed to complete the task.",[44,4709,675],{"id":674},[677,4711,4713],{"id":4712},"_12-messaging-channels","12+ Messaging Channels",[12,4715,4716],{},"OpenClaw supports a wide range of channels out of the box:",[112,4718,4719,4729,4739,4744,4757],{},[115,4720,4721,35,4723,35,4725,4728],{},[28,4722,4354],{},[28,4724,4360],{},[28,4726,4727],{},"Signal"," — personal messaging",[115,4730,4731,35,4733,35,4735,4738],{},[28,4732,4371],{},[28,4734,4377],{},[28,4736,4737],{},"Google Chat"," — workplace tools",[115,4740,4741,4743],{},[28,4742,3205],{}," — communities and gaming",[115,4745,4746,4749,4750,35,4753,4756],{},[28,4747,4748],{},"iMessage"," (via BlueBubbles), ",[28,4751,4752],{},"Matrix",[28,4754,4755],{},"Zalo"," — additional options",[115,4758,4759,4761],{},[28,4760,4384],{}," — built-in browser-based interface",[12,4763,4764],{},"You can connect multiple channels simultaneously, and even route different channels to isolated agent sessions using multi-agent routing.",[677,4766,4768],{"id":4767},"bring-your-own-llm","Bring Your Own LLM",[12,4770,4771],{},"OpenClaw isn't locked to a single AI provider. It supports:",[112,4773,4774,4780],{},[115,4775,4776,4779],{},[28,4777,4778],{},"Anthropic"," (Claude) — recommended for its long-context strength",[115,4781,4782,4785],{},[28,4783,4784],{},"OpenAI"," (GPT-4o, o1, etc.) — widely available with broad tool support",[12,4787,4788],{},"You authenticate with your own API key or OAuth credentials. You can configure model selection and failover, so if one provider has an outage, your assistant switches to a backup automatically.",[677,4790,4792],{"id":4791},"skills-and-clawhub","Skills and ClawHub",[12,4794,4795,4796,4798,4799,4802],{},"OpenClaw uses a plugin system called ",[28,4797,4274],{}," — modular extensions that teach the agent how to use specific tools. Each skill is a directory containing a ",[231,4800,4801],{},"SKILL.md"," file with YAML frontmatter and natural-language instructions.",[12,4804,4805,4806,4808],{},"The public skill registry, ",[28,4807,4315],{},", hosts over 5,700 community-built skills covering:",[112,4810,4811,4814,4817,4820,4823],{},[115,4812,4813],{},"Productivity (Google Calendar, Todoist, Notion)",[115,4815,4816],{},"Development (GitHub, CI/CD, code review)",[115,4818,4819],{},"Smart home (Home Assistant, IoT devices)",[115,4821,4822],{},"Communication (email drafting, auto-replies)",[115,4824,4825],{},"Research (web scraping, data synthesis)",[12,4827,4828],{},"Installing a skill is a single command:",[642,4830,4832],{"className":644,"code":4831,"language":646,"meta":489,"style":489},"openclaw skill install google-calendar\n",[231,4833,4834],{"__ignoreMap":489},[650,4835,4836,4839,4842,4845],{"class":652,"line":653},[650,4837,4838],{"class":656},"openclaw",[650,4840,4841],{"class":660}," skill",[650,4843,4844],{"class":660}," install",[650,4846,4847],{"class":660}," google-calendar\n",[12,4849,4850,4851,4854],{},"All skills uploaded to ClawHub are scanned with ",[28,4852,4853],{},"VirusTotal"," threat intelligence — skills flagged as suspicious are marked with warnings, while verified skills are auto-approved.",[677,4856,4858],{"id":4857},"voice-and-visual-capabilities","Voice and Visual Capabilities",[12,4860,4861],{},"OpenClaw includes built-in voice support powered by ElevenLabs:",[112,4863,4864,4870],{},[115,4865,4866,4869],{},[28,4867,4868],{},"Voice Wake"," — always-on speech activation",[115,4871,4872,4875],{},[28,4873,4874],{},"Talk Mode"," — natural conversation with voice input and output",[12,4877,4878,4881],{},[28,4879,4880],{},"Live Canvas"," provides an agent-driven visual workspace for tasks that benefit from a graphical interface, like diagram generation or data visualization.",[677,4883,4885],{"id":4884},"browser-control","Browser Control",[12,4887,4888],{},"OpenClaw can launch and control a dedicated Chrome/Chromium instance. This lets it fill out forms, take screenshots, extract data from web pages, and navigate sites autonomously — useful for tasks like booking appointments, checking prices, or monitoring dashboards.",[677,4890,4892],{"id":4891},"persistent-memory","Persistent Memory",[12,4894,4895],{},"Unlike stateless chatbots, OpenClaw maintains context across conversations. It remembers your preferences, past discussions, and accumulated knowledge. This makes it genuinely useful as a long-running assistant that improves over time.",[44,4897,4899],{"id":4898},"supported-platforms","Supported Platforms",[12,4901,4902],{},"OpenClaw runs across the major operating systems:",[795,4904,4905,4915],{},[798,4906,4907],{},[801,4908,4909,4912],{},[804,4910,4911],{},"Platform",[804,4913,4914],{},"Features",[817,4916,4917,4927,4937,4947,4957],{},[801,4918,4919,4924],{},[822,4920,4921],{},[28,4922,4923],{},"macOS",[822,4925,4926],{},"Menu bar app, full gateway, voice, browser control",[801,4928,4929,4934],{},[822,4930,4931],{},[28,4932,4933],{},"Linux",[822,4935,4936],{},"Full gateway, daemon mode, Docker support",[801,4938,4939,4944],{},[822,4940,4941],{},[28,4942,4943],{},"Windows",[822,4945,4946],{},"Via WSL2, full gateway capabilities",[801,4948,4949,4954],{},[822,4950,4951],{},[28,4952,4953],{},"iOS",[822,4955,4956],{},"Device node with Canvas, voice, camera access",[801,4958,4959,4964],{},[822,4960,4961],{},[28,4962,4963],{},"Android",[822,4965,4966],{},"Device node with camera and screen capture",[12,4968,4969,4970,4972],{},"The gateway itself requires ",[28,4971,3366],{}," and can be installed via npm, pnpm, or bun.",[44,4974,4976],{"id":4975},"installation","Installation",[12,4978,4979],{},"Getting OpenClaw running takes a few minutes:",[642,4981,4983],{"className":644,"code":4982,"language":646,"meta":489,"style":489},"# macOS / Linux\ncurl -fsSL https://openclaw.ai/install.sh | bash\n\n# Windows (PowerShell)\niwr -useb https://openclaw.ai/install.ps1 | iex\n\n# Or via npm\nnpm install -g openclaw@latest\n",[231,4984,4985,4990,5004,5008,5013,5029,5033,5038],{"__ignoreMap":489},[650,4986,4987],{"class":652,"line":653},[650,4988,4989],{"class":930},"# macOS / Linux\n",[650,4991,4992,4994,4996,4999,5001],{"class":652,"line":490},[650,4993,1765],{"class":656},[650,4995,1768],{"class":664},[650,4997,4998],{"class":660}," https://openclaw.ai/install.sh",[650,5000,1775],{"class":1774},[650,5002,5003],{"class":656}," bash\n",[650,5005,5006],{"class":652,"line":944},[650,5007,989],{"emptyLinePlaceholder":510},[650,5009,5010],{"class":652,"line":956},[650,5011,5012],{"class":930},"# Windows (PowerShell)\n",[650,5014,5015,5018,5021,5024,5026],{"class":652,"line":964},[650,5016,5017],{"class":656},"iwr",[650,5019,5020],{"class":664}," -useb",[650,5022,5023],{"class":660}," https://openclaw.ai/install.ps1",[650,5025,1775],{"class":1774},[650,5027,5028],{"class":656}," iex\n",[650,5030,5031],{"class":652,"line":975},[650,5032,989],{"emptyLinePlaceholder":510},[650,5034,5035],{"class":652,"line":986},[650,5036,5037],{"class":930},"# Or via npm\n",[650,5039,5040,5043,5045,5048],{"class":652,"line":992},[650,5041,5042],{"class":656},"npm",[650,5044,4844],{"class":660},[650,5046,5047],{"class":664}," -g",[650,5049,5050],{"class":660}," openclaw@latest\n",[12,5052,5053],{},"Then run the onboarding wizard:",[642,5055,5057],{"className":644,"code":5056,"language":646,"meta":489,"style":489},"openclaw onboard --install-daemon\n",[231,5058,5059],{"__ignoreMap":489},[650,5060,5061,5063,5066],{"class":652,"line":653},[650,5062,4838],{"class":656},[650,5064,5065],{"class":660}," onboard",[650,5067,5068],{"class":664}," --install-daemon\n",[12,5070,5071,5072,157],{},"This walks you through gateway setup, channel configuration, and skill installation. Once complete, access the dashboard at ",[231,5073,5074],{},"http://127.0.0.1:18789/",[44,5076,5078],{"id":5077},"self-hosting-vs-managed-hosting","Self-Hosting vs Managed Hosting",[12,5080,5081],{},"Running OpenClaw yourself means managing a server 24/7 — security patches, Docker configuration, SSL certificates, backups, and uptime monitoring. Minimum requirements for a VPS are 2 CPU cores, 2 GB RAM, and 20 GB SSD storage.",[12,5083,5084,5085,5088,5089,5091],{},"For users who want OpenClaw running around the clock without the operational overhead, ",[28,5086,5087],{},"managed hosting"," handles all of that. ",[1584,5090,295],{"href":1601}," provides fully managed OpenClaw instances with automatic backups, a web dashboard, EU data residency, and deployment in minutes rather than hours.",[12,5093,5094,5095,157],{},"Read our detailed comparison: ",[1584,5096,5097],{"href":4088},"Managed Hosting vs Self-Hosting",[44,5099,5101],{"id":5100},"who-uses-openclaw","Who Uses OpenClaw?",[12,5103,5104],{},"OpenClaw powers a wide range of use cases:",[112,5106,5107,5113,5119,5125,5131],{},[115,5108,5109,5112],{},[28,5110,5111],{},"Personal assistants"," — daily planning, email triage, knowledge management",[115,5114,5115,5118],{},[28,5116,5117],{},"Customer support"," — first-line inquiries, ticket routing, FAQ handling",[115,5120,5121,5124],{},[28,5122,5123],{},"Development"," — code review, documentation generation, CI/CD automation",[115,5126,5127,5130],{},[28,5128,5129],{},"Research"," — web monitoring, data synthesis, trend analysis",[115,5132,5133,5136],{},[28,5134,5135],{},"Community management"," — Discord AI assistants, onboarding flows, moderation",[12,5138,5139,5140,5144],{},"See our full list of ",[1584,5141,5143],{"href":5142},"/use-cases","use cases"," for more examples.",[44,5146,3154],{"id":3153},[12,5148,5149,5150,5152,5153,5157],{},"If you want to try OpenClaw without the complexity of self-hosting, ClawNest offers a ",[1584,5151,3219],{"href":1605}," — no credit card required. Follow our ",[1584,5154,5156],{"href":5155},"/tutorial","step-by-step tutorial"," to deploy your first AI assistant in under 5 minutes.",[1609,5159,5160],{},"html pre.shiki code .sScJk, html code.shiki .sScJk{--shiki-default:#6F42C1;--shiki-dark:#B392F0}html pre.shiki code .sZZnC, html code.shiki .sZZnC{--shiki-default:#032F62;--shiki-dark:#9ECBFF}html .default .shiki span {color: var(--shiki-default);background: var(--shiki-default-bg);font-style: var(--shiki-default-font-style);font-weight: var(--shiki-default-font-weight);text-decoration: var(--shiki-default-text-decoration);}html .shiki span {color: var(--shiki-default);background: var(--shiki-default-bg);font-style: var(--shiki-default-font-style);font-weight: var(--shiki-default-font-weight);text-decoration: var(--shiki-default-text-decoration);}html .dark .shiki span {color: var(--shiki-dark);background: var(--shiki-dark-bg);font-style: var(--shiki-dark-font-style);font-weight: var(--shiki-dark-font-weight);text-decoration: var(--shiki-dark-text-decoration);}html.dark .shiki span {color: var(--shiki-dark);background: var(--shiki-dark-bg);font-style: var(--shiki-dark-font-style);font-weight: var(--shiki-dark-font-weight);text-decoration: var(--shiki-dark-text-decoration);}html pre.shiki code .sJ8bj, html code.shiki .sJ8bj{--shiki-default:#6A737D;--shiki-dark:#6A737D}html pre.shiki code .sj4cs, html code.shiki .sj4cs{--shiki-default:#005CC5;--shiki-dark:#79B8FF}html pre.shiki code .szBVR, html code.shiki .szBVR{--shiki-default:#D73A49;--shiki-dark:#F97583}",{"title":489,"searchDepth":490,"depth":490,"links":5162},[5163,5164,5165,5173,5174,5175,5176,5177],{"id":4665,"depth":490,"text":4666},{"id":4690,"depth":490,"text":4691},{"id":674,"depth":490,"text":675,"children":5166},[5167,5168,5169,5170,5171,5172],{"id":4712,"depth":944,"text":4713},{"id":4767,"depth":944,"text":4768},{"id":4791,"depth":944,"text":4792},{"id":4857,"depth":944,"text":4858},{"id":4884,"depth":944,"text":4885},{"id":4891,"depth":944,"text":4892},{"id":4898,"depth":490,"text":4899},{"id":4975,"depth":490,"text":4976},{"id":5077,"depth":490,"text":5078},{"id":5100,"depth":490,"text":5101},{"id":3153,"depth":490,"text":3154},"2026-02-10","Learn what OpenClaw is, how it works, and why 60,000+ developers use this open-source personal AI assistant. Covers architecture, skills, channels, and hosting options.",{},"/blog/what-is-openclaw",{"title":4660,"description":5179},"blog/what-is-openclaw",[68,3259,5185],"AI Assistant","iwI8XeRcUceD6OJkoe1aF-Ki33Mx7eB8lccvDpsG-H8",{"id":5188,"title":5189,"author":7,"body":5190,"date":5599,"description":5600,"draft":506,"extension":507,"image":4649,"meta":5601,"navigation":510,"path":5602,"seo":5603,"stem":5604,"tags":5605,"__hash__":5606},"blog/blog/openclaw-hosting-options-compared.md","OpenClaw Hosting Options Compared: VPS vs Mac Mini vs Managed",{"type":9,"value":5191,"toc":5576},[5192,5196,5199,5203,5205,5208,5212,5231,5235,5267,5271,5274,5278,5281,5284,5287,5307,5310,5342,5345,5348,5352,5355,5361,5364,5399,5402,5422,5425,5428,5432,5545,5549,5569],[44,5193,5195],{"id":5194},"three-ways-to-run-openclaw","Three Ways to Run OpenClaw",[12,5197,5198],{},"When you decide to run OpenClaw, you have three main hosting options. Each comes with different trade-offs in terms of cost, complexity, control, and reliability.",[44,5200,5202],{"id":5201},"option-1-self-host-on-a-vps","Option 1: Self-Host on a VPS",[677,5204,2909],{"id":2908},[12,5206,5207],{},"You rent a virtual private server from a provider like Hetzner, DigitalOcean, or AWS. You install the operating system, set up Docker or a runtime environment, install OpenClaw, configure networking, and manage everything yourself.",[677,5209,5211],{"id":5210},"pros","Pros",[112,5213,5214,5219,5225],{},[115,5215,5216,5218],{},[28,5217,3893],{}," — complete control over the server",[115,5220,5221,5224],{},[28,5222,5223],{},"Low monthly cost"," — VPS plans start around €5–20/mo",[115,5226,5227,5230],{},[28,5228,5229],{},"Provider choice"," — pick any data center location",[677,5232,5234],{"id":5233},"cons","Cons",[112,5236,5237,5243,5249,5255,5261],{},[115,5238,5239,5242],{},[28,5240,5241],{},"Setup time"," — expect hours to days for initial configuration",[115,5244,5245,5248],{},[28,5246,5247],{},"Ongoing maintenance"," — OS updates, security patches, monitoring",[115,5250,5251,5254],{},[28,5252,5253],{},"Manual backups"," — you're responsible for backup scripts and storage",[115,5256,5257,5260],{},[28,5258,5259],{},"Requires Linux knowledge"," — SSH, firewall rules, process management",[115,5262,5263,5266],{},[28,5264,5265],{},"No GUI by default"," — command-line only unless you set up remote desktop",[677,5268,5270],{"id":5269},"best-for","Best For",[12,5272,5273],{},"Experienced Linux users who want maximum control and don't mind investing time in server management.",[44,5275,5277],{"id":5276},"option-2-run-locally-on-a-mac-mini","Option 2: Run Locally on a Mac Mini",[677,5279,2909],{"id":5280},"how-it-works-1",[12,5282,5283],{},"You buy a Mac Mini (or use an existing Mac), install OpenClaw locally, and keep the machine running 24/7 at home or in an office.",[677,5285,5211],{"id":5286},"pros-1",[112,5288,5289,5295,5301],{},[115,5290,5291,5294],{},[28,5292,5293],{},"No monthly hosting cost"," — just electricity",[115,5296,5297,5300],{},[28,5298,5299],{},"Physical access"," — the hardware is right there",[115,5302,5303,5306],{},[28,5304,5305],{},"Native macOS GUI"," — easy graphical interface access",[677,5308,5234],{"id":5309},"cons-1",[112,5311,5312,5318,5324,5330,5336],{},[115,5313,5314,5317],{},[28,5315,5316],{},"Upfront hardware cost"," — €700+ for a Mac Mini",[115,5319,5320,5323],{},[28,5321,5322],{},"Home internet dependency"," — downtime if your ISP has issues",[115,5325,5326,5329],{},[28,5327,5328],{},"Power and cooling"," — running 24/7 adds to your electricity bill",[115,5331,5332,5335],{},[28,5333,5334],{},"No geographic redundancy"," — single point of failure",[115,5337,5338,5341],{},[28,5339,5340],{},"Manual everything"," — updates, backups, security are all on you",[677,5343,5270],{"id":5344},"best-for-1",[12,5346,5347],{},"Hobbyists who already own a Mac and want to experiment without monthly costs.",[44,5349,5351],{"id":5350},"option-3-managed-hosting-with-clawnest","Option 3: Managed Hosting with ClawNest",[677,5353,2909],{"id":5354},"how-it-works-2",[12,5356,5357,5358,5360],{},"You sign up at ",[1584,5359,295],{"href":1605},", configure your AI assistant through a web dashboard, and click deploy. ClawNest handles the servers, updates, backups, and monitoring.",[677,5362,5211],{"id":5363},"pros-2",[112,5365,5366,5372,5377,5382,5388,5393],{},[115,5367,5368,5371],{},[28,5369,5370],{},"Deploy in minutes"," — no server setup required",[115,5373,5374,5376],{},[28,5375,3953],{}," — with restore and rollback tooling",[115,5378,5379,3972],{},[28,5380,5381],{},"Web dashboard + GUI access",[115,5383,5384,5387],{},[28,5385,5386],{},"24/7 uptime"," — cloud infrastructure with auto-scaling",[115,5389,5390,5392],{},[28,5391,3965],{}," — hosted in Sweden",[115,5394,5395,5398],{},[28,5396,5397],{},"BYOK pricing"," — bring your own API key, no AI markup",[677,5400,5234],{"id":5401},"cons-2",[112,5403,5404,5410,5416],{},[115,5405,5406,5409],{},[28,5407,5408],{},"Monthly cost"," — plans start at €12/mo",[115,5411,5412,5415],{},[28,5413,5414],{},"Less low-level control"," — shell access available on Pro+ plans",[115,5417,5418,5421],{},[28,5419,5420],{},"Vendor dependency"," — you rely on ClawNest infrastructure",[677,5423,5270],{"id":5424},"best-for-2",[12,5426,5427],{},"Anyone who wants reliable, always-on OpenClaw AI assistants without managing servers — from hobbyists to teams.",[44,5429,5431],{"id":5430},"quick-comparison-table","Quick Comparison Table",[795,5433,5434,5448],{},[798,5435,5436],{},[801,5437,5438,5440,5443,5446],{},[804,5439,1250],{},[804,5441,5442],{},"VPS (DIY)",[804,5444,5445],{},"Mac Mini",[804,5447,295],{},[817,5449,5450,5463,5476,5489,5500,5510,5521,5534],{},[801,5451,5452,5454,5457,5460],{},[822,5453,5241],{},[822,5455,5456],{},"Hours–days",[822,5458,5459],{},"Hours",[822,5461,5462],{},"Minutes",[801,5464,5465,5467,5470,5473],{},[822,5466,5408],{},[822,5468,5469],{},"€5–20 + time",[822,5471,5472],{},"Electricity",[822,5474,5475],{},"From €12",[801,5477,5478,5481,5484,5487],{},[822,5479,5480],{},"Upfront cost",[822,5482,5483],{},"None",[822,5485,5486],{},"€700+",[822,5488,5483],{},[801,5490,5491,5494,5496,5498],{},[822,5492,5493],{},"Server management",[822,5495,3566],{},[822,5497,3566],{},[822,5499,3569],{},[801,5501,5502,5504,5506,5508],{},[822,5503,3620],{},[822,5505,1286],{},[822,5507,1286],{},[822,5509,3579],{},[801,5511,5512,5514,5517,5519],{},[822,5513,4408],{},[822,5515,5516],{},"Manual setup",[822,5518,1372],{},[822,5520,3779],{},[801,5522,5523,5526,5529,5531],{},[822,5524,5525],{},"Shell access",[822,5527,5528],{},"Full root",[822,5530,4471],{},[822,5532,5533],{},"Pro plan+",[801,5535,5536,5538,5540,5543],{},[822,5537,4482],{},[822,5539,1322],{},[822,5541,5542],{},"Your home",[822,5544,4485],{},[44,5546,5548],{"id":5547},"which-should-you-choose","Which Should You Choose?",[112,5550,5551,5557,5563],{},[115,5552,5553,5556],{},[28,5554,5555],{},"Choose VPS"," if you're a Linux expert who wants full control and enjoys server management.",[115,5558,5559,5562],{},[28,5560,5561],{},"Choose Mac Mini"," if you already own the hardware and are fine with a local-only setup.",[115,5564,5565,5568],{},[28,5566,5567],{},"Choose ClawNest"," if you want OpenClaw running 24/7 with minimal effort and professional infrastructure.",[12,5570,5571,5572,157],{},"For a more detailed feature-by-feature comparison, visit our ",[1584,5573,5575],{"href":5574},"/compare","comparison page",{"title":489,"searchDepth":490,"depth":490,"links":5577},[5578,5579,5585,5591,5597,5598],{"id":5194,"depth":490,"text":5195},{"id":5201,"depth":490,"text":5202,"children":5580},[5581,5582,5583,5584],{"id":2908,"depth":944,"text":2909},{"id":5210,"depth":944,"text":5211},{"id":5233,"depth":944,"text":5234},{"id":5269,"depth":944,"text":5270},{"id":5276,"depth":490,"text":5277,"children":5586},[5587,5588,5589,5590],{"id":5280,"depth":944,"text":2909},{"id":5286,"depth":944,"text":5211},{"id":5309,"depth":944,"text":5234},{"id":5344,"depth":944,"text":5270},{"id":5350,"depth":490,"text":5351,"children":5592},[5593,5594,5595,5596],{"id":5354,"depth":944,"text":2909},{"id":5363,"depth":944,"text":5211},{"id":5401,"depth":944,"text":5234},{"id":5424,"depth":944,"text":5270},{"id":5430,"depth":490,"text":5431},{"id":5547,"depth":490,"text":5548},"2026-02-08","Compare the three main ways to host OpenClaw — self-hosted VPS, local Mac Mini, and managed hosting — to find the best option for your needs.",{},"/blog/openclaw-hosting-options-compared",{"title":5189,"description":5600},"blog/openclaw-hosting-options-compared",[4092,4093,3259],"uTaTRFcklK5PwK8l1evTU1fBaZ1Uq_opGQoOtYlt3oc",{"id":5608,"title":5609,"author":7,"body":5610,"date":5819,"description":5820,"draft":506,"extension":507,"image":4649,"meta":5821,"navigation":510,"path":5822,"seo":5823,"stem":5824,"tags":5825,"__hash__":5826},"blog/blog/deploy-openclaw-bot-5-minutes.md","How to Deploy an OpenClaw AI Assistant in 5 Minutes with ClawNest",{"type":9,"value":5611,"toc":5804},[5612,5616,5619,5623,5626,5637,5640,5644,5650,5654,5656,5671,5676,5680,5683,5696,5700,5703,5721,5724,5726,5731,5741,5745,5748,5773,5775,5777,5780,5782,5788,5790,5793,5795],[44,5613,5615],{"id":5614},"deploy-your-first-ai-assistant-in-5-minutes","Deploy Your First AI Assistant in 5 Minutes",[12,5617,5618],{},"Getting an OpenClaw AI assistant running usually involves setting up servers, installing dependencies, and configuring networking. With ClawNest, you skip all of that. Here's how to go from zero to a running AI assistant in about 5 minutes.",[44,5620,5622],{"id":5621},"prerequisites","Prerequisites",[12,5624,5625],{},"Before you start, you'll need:",[112,5627,5628,5632],{},[115,5629,5630,4124],{},[28,5631,4123],{},[115,5633,5634,5636],{},[28,5635,4129],{}," from an LLM provider (OpenAI, Anthropic, or OpenRouter)",[12,5638,5639],{},"That's it. No server, no Docker, no SSH keys.",[44,5641,5643],{"id":5642},"step-1-create-your-account","Step 1: Create Your Account",[12,5645,4197,5646,5649],{},[1584,5647,3667],{"href":3214,"rel":5648},[1588]," and sign up. The free trial gives you 3 days of full access with one AI assistant — no credit card required.",[44,5651,5653],{"id":5652},"step-2-get-your-api-key","Step 2: Get Your API Key",[12,5655,4140],{},[1547,5657,5658,5665,5668],{},[115,5659,5660,5661,5664],{},"Go to your preferred provider's website (e.g., ",[1584,5662,4178],{"href":4176,"rel":5663},[1588],")",[115,5666,5667],{},"Create a new API key",[115,5669,5670],{},"Copy it to your clipboard",[12,5672,4187,5673,5675],{},[28,5674,3878],{}," — you pay providers directly at their rates. We add zero markup on AI tokens.",[44,5677,5679],{"id":5678},"step-3-create-a-new-ai-assistant","Step 3: Create a New AI Assistant",[12,5681,5682],{},"From the ClawNest dashboard:",[1547,5684,5685,5689,5692,5694],{},[115,5686,4220,5687],{},[28,5688,4223],{},[115,5690,5691],{},"Give your AI assistant a name",[115,5693,4229],{},[115,5695,4232],{},[44,5697,5699],{"id":5698},"step-4-configure-your-ai-assistant","Step 4: Configure Your AI Assistant",[12,5701,5702],{},"Use the visual editor to set up your AI assistant:",[112,5704,5705,5710,5715],{},[115,5706,5707,5709],{},[28,5708,4246],{}," — define how your AI assistant should behave and respond",[115,5711,5712,5714],{},[28,5713,4274],{}," — enable plugins like web search, file handling, or code execution",[115,5716,5717,5720],{},[28,5718,5719],{},"Rules"," — set boundaries and safety guardrails",[12,5722,5723],{},"No coding required. Everything is configured through the dashboard UI.",[44,5725,4389],{"id":4388},[12,5727,4220,5728,5730],{},[28,5729,4394],{}," and your AI assistant goes live. It's now running 24/7 on ClawNest cloud infrastructure with:",[112,5732,5733,5735,5738],{},[115,5734,3953],{},[115,5736,5737],{},"GUI access for further configuration",[115,5739,5740],{},"Status monitoring from the dashboard",[44,5742,5744],{"id":5743},"whats-next","What's Next?",[12,5746,5747],{},"Once your AI assistant is running, you can:",[112,5749,5750,5756,5762,5767],{},[115,5751,5752,5755],{},[28,5753,5754],{},"Fine-tune its personality"," through the GUI interface",[115,5757,5758,5761],{},[28,5759,5760],{},"Add more skills"," to expand its capabilities",[115,5763,5764,4415],{},[28,5765,5766],{},"Monitor usage",[115,5768,5769,5772],{},[28,5770,5771],{},"Add more AI assistants"," if you need them (from €8/mo each as an add-on)",[44,5774,3495],{"id":4576},[677,5776,4580],{"id":4579},[12,5778,5779],{},"Make sure your API key is active and has sufficient credits with your provider. You can test the connection from the ClawNest dashboard.",[677,5781,4587],{"id":4586},[12,5783,5784,5785,157],{},"Check the AI assistant status in your dashboard. If it shows as offline, try redeploying. For persistent issues, reach out on our ",[1584,5786,3205],{"href":3203,"rel":5787},[1588],[677,5789,4594],{"id":4593},[12,5791,5792],{},"Our primary support channel is Discord. The community and team are active and happy to help with setup questions.",[1591,5794],{},[12,5796,5797,5798,5801,5802,157],{},"For the full step-by-step tutorial with screenshots, visit our ",[1584,5799,5800],{"href":5155},"tutorial page",". To explore all the features available on your plan, check out the ",[1584,5803,4619],{"href":4618},{"title":489,"searchDepth":490,"depth":490,"links":5805},[5806,5807,5808,5809,5810,5811,5812,5813,5814],{"id":5614,"depth":490,"text":5615},{"id":5621,"depth":490,"text":5622},{"id":5642,"depth":490,"text":5643},{"id":5652,"depth":490,"text":5653},{"id":5678,"depth":490,"text":5679},{"id":5698,"depth":490,"text":5699},{"id":4388,"depth":490,"text":4389},{"id":5743,"depth":490,"text":5744},{"id":4576,"depth":490,"text":3495,"children":5815},[5816,5817,5818],{"id":4579,"depth":944,"text":4580},{"id":4586,"depth":944,"text":4587},{"id":4593,"depth":944,"text":4594},"2026-02-06","Step-by-step guide to deploying your first OpenClaw AI assistant on ClawNest managed hosting — from sign-up to a running AI assistant in under 5 minutes.",{},"/blog/deploy-openclaw-bot-5-minutes",{"title":5609,"description":5820},"blog/deploy-openclaw-bot-5-minutes",[4655,4656,3154],"iPs_GhUi_q_GgpXeCEDVoFxPsDvvb1Y7eEX8PvUqklY",{"id":5828,"title":5829,"author":7,"body":5830,"date":6096,"description":6097,"draft":506,"extension":507,"image":4649,"meta":6098,"navigation":510,"path":6099,"seo":6100,"stem":6101,"tags":6102,"__hash__":6105},"blog/blog/openclaw-use-cases-2026.md","OpenClaw Use Cases in 2026: What People Are Building with AI Assistants",{"type":9,"value":5831,"toc":6084},[5832,5836,5839,5842,5846,5853,5867,5870,5874,5877,5903,5909,5913,5916,5930,5933,5937,5940,5954,5957,5961,5964,5990,5994,5997,6011,6015,6018,6032,6036,6039,6053,6057,6060,6074],[44,5833,5835],{"id":5834},"what-are-people-building-with-openclaw","What Are People Building with OpenClaw?",[12,5837,5838],{},"OpenClaw has grown from a niche tool into one of the most flexible open-source AI engines available. In 2026, users are deploying OpenClaw AI assistants for everything from personal productivity to enterprise customer support.",[12,5840,5841],{},"Here's a look at the most popular use cases.",[44,5843,5845],{"id":5844},"personal-ai-assistants","Personal AI Assistants",[12,5847,5848,5849,5852],{},"The most common use case is the ",[28,5850,5851],{},"always-on personal assistant",". Unlike ChatGPT or Claude, an OpenClaw AI assistant:",[112,5854,5855,5858,5861,5864],{},[115,5856,5857],{},"Remembers everything across sessions (persistent memory)",[115,5859,5860],{},"Runs 24/7 without needing a browser tab open",[115,5862,5863],{},"Can be customized with specific skills and behaviors",[115,5865,5866],{},"Works with your preferred LLM provider",[12,5868,5869],{},"Users configure their assistants for daily planning, email drafting, knowledge management, and quick lookups.",[44,5871,5873],{"id":5872},"customer-support-ai-assistants","Customer Support AI Assistants",[12,5875,5876],{},"Small businesses and SaaS companies use OpenClaw to handle first-line customer support:",[112,5878,5879,5885,5891,5897],{},[115,5880,5881,5884],{},[28,5882,5883],{},"FAQ handling"," — answer common questions instantly",[115,5886,5887,5890],{},[28,5888,5889],{},"Ticket triage"," — categorize and route support requests",[115,5892,5893,5896],{},[28,5894,5895],{},"Product guidance"," — walk users through features",[115,5898,5899,5902],{},[28,5900,5901],{},"Order tracking"," — provide status updates from connected APIs",[12,5904,5905,5906,5908],{},"With managed hosting on ",[1584,5907,295],{"href":1605},", support AI assistants run 24/7 with automatic backups and no server management.",[44,5910,5912],{"id":5911},"creative-writing-partners","Creative Writing Partners",[12,5914,5915],{},"Writers, content creators, and marketing teams use OpenClaw as a creative collaborator:",[112,5917,5918,5921,5924,5927],{},[115,5919,5920],{},"Brainstorm ideas and outlines",[115,5922,5923],{},"Draft blog posts, social media content, and marketing copy",[115,5925,5926],{},"Edit and refine existing text",[115,5928,5929],{},"Maintain consistent brand voice across content",[12,5931,5932],{},"The persistent memory feature means the AI assistant learns your style and preferences over time.",[44,5934,5936],{"id":5935},"code-assistants","Code Assistants",[12,5938,5939],{},"Developers deploy OpenClaw AI assistants as coding companions that:",[112,5941,5942,5945,5948,5951],{},[115,5943,5944],{},"Review pull requests and suggest improvements",[115,5946,5947],{},"Help debug issues with context from previous sessions",[115,5949,5950],{},"Generate documentation from codebases",[115,5952,5953],{},"Assist with learning new frameworks and languages",[12,5955,5956],{},"Since OpenClaw supports multiple LLM providers, developers can choose the model best suited for code tasks.",[44,5958,5960],{"id":5959},"ai-tutors","AI Tutors",[12,5962,5963],{},"Educators and learners use OpenClaw to create patient, always-available tutoring AI assistants:",[112,5965,5966,5972,5978,5984],{},[115,5967,5968,5971],{},[28,5969,5970],{},"Language learning"," — conversation practice with correction",[115,5973,5974,5977],{},[28,5975,5976],{},"Math tutoring"," — step-by-step problem solving",[115,5979,5980,5983],{},[28,5981,5982],{},"Exam preparation"," — question banks and explanations",[115,5985,5986,5989],{},[28,5987,5988],{},"Skill development"," — guided learning paths",[44,5991,5993],{"id":5992},"community-management","Community Management",[12,5995,5996],{},"Discord servers and online communities use OpenClaw AI assistants for:",[112,5998,5999,6002,6005,6008],{},[115,6000,6001],{},"Welcome messages and onboarding flows",[115,6003,6004],{},"FAQ responses and knowledge base access",[115,6006,6007],{},"Moderation assistance",[115,6009,6010],{},"Event coordination and reminders",[44,6012,6014],{"id":6013},"workflow-automation","Workflow Automation",[12,6016,6017],{},"Power users combine OpenClaw's skills system with external APIs to automate:",[112,6019,6020,6023,6026,6029],{},[115,6021,6022],{},"Data processing and reporting",[115,6024,6025],{},"Email workflows and responses",[115,6027,6028],{},"Task routing and notifications",[115,6030,6031],{},"Scheduled content generation",[44,6033,6035],{"id":6034},"research-and-analysis","Research and Analysis",[12,6037,6038],{},"Researchers and analysts deploy AI assistants that:",[112,6040,6041,6044,6047,6050],{},[115,6042,6043],{},"Gather information from multiple sources",[115,6045,6046],{},"Synthesize findings into summaries",[115,6048,6049],{},"Monitor trends and competitors",[115,6051,6052],{},"Generate structured reports",[44,6054,6056],{"id":6055},"getting-started-with-your-use-case","Getting Started with Your Use Case",[12,6058,6059],{},"No matter what you want to build, the process is the same:",[1547,6061,6062,6068,6071],{},[115,6063,6064,6067],{},[1584,6065,6066],{"href":1605},"Sign up for ClawNest"," (free 3-day trial)",[115,6069,6070],{},"Configure your AI assistant's personality and skills",[115,6072,6073],{},"Deploy and start using it",[12,6075,6076,6077,6080,6081,157],{},"For detailed inspiration, visit our ",[1584,6078,6079],{"href":5142},"use cases page",". For a step-by-step setup guide, check out the ",[1584,6082,6083],{"href":5155},"tutorial",{"title":489,"searchDepth":490,"depth":490,"links":6085},[6086,6087,6088,6089,6090,6091,6092,6093,6094,6095],{"id":5834,"depth":490,"text":5835},{"id":5844,"depth":490,"text":5845},{"id":5872,"depth":490,"text":5873},{"id":5911,"depth":490,"text":5912},{"id":5935,"depth":490,"text":5936},{"id":5959,"depth":490,"text":5960},{"id":5992,"depth":490,"text":5993},{"id":6013,"depth":490,"text":6014},{"id":6034,"depth":490,"text":6035},{"id":6055,"depth":490,"text":6056},"2026-02-04","Discover the most popular OpenClaw use cases in 2026 — from personal assistants and customer support AI assistants to creative writing partners and research analysts.",{},"/blog/openclaw-use-cases-2026",{"title":5829,"description":6097},"blog/openclaw-use-cases-2026",[6103,3816,6104],"Use Cases","2026","z3UE7yLiaMhFVWVYJ_48m4mSaSipIXwf5ngwGtL76xI",{"id":6107,"title":6108,"author":7,"body":6109,"date":6450,"description":6451,"draft":506,"extension":507,"image":4649,"meta":6452,"navigation":510,"path":6453,"seo":6454,"stem":6455,"tags":6456,"__hash__":6460},"blog/blog/byok-ai-hosting-explained.md","BYOK AI Hosting Explained: Why Bring Your Own Key Matters",{"type":9,"value":6110,"toc":6433},[6111,6115,6122,6126,6129,6143,6146,6150,6153,6167,6174,6178,6182,6185,6189,6192,6218,6222,6225,6229,6232,6236,6275,6279,6317,6320,6324,6327,6352,6362,6366,6411,6414,6418,6428],[44,6112,6114],{"id":6113},"what-is-byok","What Is BYOK?",[12,6116,6117,6118,6121],{},"BYOK stands for ",[28,6119,6120],{},"Bring Your Own Key",". In the context of AI hosting, it means you provide your own API keys from LLM providers (like OpenAI, Anthropic, or OpenRouter) rather than paying the hosting platform for AI tokens.",[44,6123,6125],{"id":6124},"how-most-ai-platforms-work","How Most AI Platforms Work",[12,6127,6128],{},"Many AI platforms bundle infrastructure and AI tokens together. When you use them, you're paying:",[1547,6130,6131,6137],{},[115,6132,6133,6136],{},[28,6134,6135],{},"Infrastructure cost"," — servers, bandwidth, storage",[115,6138,6139,6142],{},[28,6140,6141],{},"AI token markup"," — the platform buys tokens wholesale and resells them at a higher price",[12,6144,6145],{},"This markup can be significant — sometimes 2x to 5x the provider's direct pricing. And since AI usage scales with your AI assistant's activity, these hidden costs can add up fast.",[44,6147,6149],{"id":6148},"how-byok-works-at-clawnest","How BYOK Works at ClawNest",[12,6151,6152],{},"With ClawNest's BYOK model:",[1547,6154,6155,6161],{},[115,6156,6157,6160],{},[28,6158,6159],{},"You pay ClawNest"," for managed hosting (infrastructure, backups, dashboard, monitoring)",[115,6162,6163,6166],{},[28,6164,6165],{},"You pay your LLM provider directly"," for AI tokens at their published rates",[12,6168,6169,6170,6173],{},"There's ",[28,6171,6172],{},"zero markup"," on AI tokens. ClawNest never touches, resells, or marks up your API usage.",[44,6175,6177],{"id":6176},"why-byok-saves-you-money","Why BYOK Saves You Money",[677,6179,6181],{"id":6180},"transparent-pricing","Transparent Pricing",[12,6183,6184],{},"You always know exactly what you're paying for. Hosting costs are fixed and predictable. AI costs depend on your usage and are billed directly by your provider.",[677,6186,6188],{"id":6187},"provider-freedom","Provider Freedom",[12,6190,6191],{},"Not locked into one AI provider? Good. With BYOK, you can:",[112,6193,6194,6200,6206,6212],{},[115,6195,6196,6199],{},[28,6197,6198],{},"Switch providers"," anytime without changing your hosting",[115,6201,6202,6205],{},[28,6203,6204],{},"Use different models"," for different AI assistants",[115,6207,6208,6211],{},[28,6209,6210],{},"Take advantage of new pricing"," when providers lower their rates",[115,6213,6214,6217],{},[28,6215,6216],{},"Access enterprise agreements"," if you have volume deals with a provider",[677,6219,6221],{"id":6220},"no-surprise-bills","No Surprise Bills",[12,6223,6224],{},"Since AI costs go directly to your provider (where you can set spending limits), there are no unexpected charges from the hosting platform. Your ClawNest bill is the same every month.",[44,6226,6228],{"id":6227},"cost-example","Cost Example",[12,6230,6231],{},"Let's compare a hypothetical monthly bill:",[677,6233,6235],{"id":6234},"platform-with-token-markup","Platform with Token Markup",[795,6237,6238,6247],{},[798,6239,6240],{},[801,6241,6242,6244],{},[804,6243,3698],{},[804,6245,6246],{},"Cost",[817,6248,6249,6256,6264],{},[801,6250,6251,6253],{},[822,6252,4092],{},[822,6254,6255],{},"\"Free\" (bundled)",[801,6257,6258,6261],{},[822,6259,6260],{},"AI tokens (marked up 3x)",[822,6262,6263],{},"~€60",[801,6265,6266,6271],{},[822,6267,6268],{},[28,6269,6270],{},"Total",[822,6272,6273],{},[28,6274,6263],{},[677,6276,6278],{"id":6277},"clawnest-byok","ClawNest (BYOK)",[795,6280,6281,6289],{},[798,6282,6283],{},[801,6284,6285,6287],{},[804,6286,3698],{},[804,6288,6246],{},[817,6290,6291,6298,6306],{},[801,6292,6293,6296],{},[822,6294,6295],{},"Hosting (Starter plan)",[822,6297,3771],{},[801,6299,6300,6303],{},[822,6301,6302],{},"AI tokens (direct pricing)",[822,6304,6305],{},"~€20",[801,6307,6308,6312],{},[822,6309,6310],{},[28,6311,6270],{},[822,6313,6314],{},[28,6315,6316],{},"~€32",[12,6318,6319],{},"The savings grow as your usage increases. For heavy users, BYOK can cut AI costs by 50–80%.",[44,6321,6323],{"id":6322},"setting-up-byok-on-clawnest","Setting Up BYOK on ClawNest",[12,6325,6326],{},"Getting your API key is straightforward:",[1547,6328,6329,6335,6341,6347],{},[115,6330,6331,6334],{},[28,6332,6333],{},"Choose a provider"," — OpenAI, Anthropic, or OpenRouter",[115,6336,6337,6340],{},[28,6338,6339],{},"Create an API key"," on their platform",[115,6342,6343,6346],{},[28,6344,6345],{},"Paste it into ClawNest"," when creating or configuring an AI assistant",[115,6348,6349,4415],{},[28,6350,6351],{},"Test the connection",[12,6353,6354,6355,6357,6358,6361],{},"Need help? Our ",[1584,6356,6083],{"href":5155}," walks you through the entire process, and our ",[1584,6359,4493],{"href":3203,"rel":6360},[1588]," is always available for 1:1 guidance.",[44,6363,6365],{"id":6364},"which-provider-should-you-choose","Which Provider Should You Choose?",[795,6367,6368,6380],{},[798,6369,6370],{},[801,6371,6372,6375,6377],{},[804,6373,6374],{},"Provider",[804,6376,5270],{},[804,6378,6379],{},"Pricing Model",[817,6381,6382,6392,6401],{},[801,6383,6384,6386,6389],{},[822,6385,4784],{},[822,6387,6388],{},"General purpose, GPT models",[822,6390,6391],{},"Pay-per-token",[801,6393,6394,6396,6399],{},[822,6395,4778],{},[822,6397,6398],{},"Long context, safety-focused",[822,6400,6391],{},[801,6402,6403,6405,6408],{},[822,6404,1731],{},[822,6406,6407],{},"Model variety, cost optimization",[822,6409,6410],{},"Pay-per-token (multi-model)",[12,6412,6413],{},"All three work seamlessly with ClawNest. You can even use different providers for different AI assistants on the same account.",[44,6415,6417],{"id":6416},"the-bottom-line","The Bottom Line",[12,6419,6420,6421,6424,6425,157],{},"BYOK means you get ",[28,6422,6423],{},"transparent, fair pricing"," without hidden AI markups. You control your AI costs, choose your providers, and only pay ClawNest for what we actually provide: reliable, managed ",[1584,6426,6427],{"href":1601},"OpenClaw hosting",[12,6429,6430,6431,3220],{},"Ready to try it? Start with a ",[1584,6432,3219],{"href":1605},{"title":489,"searchDepth":490,"depth":490,"links":6434},[6435,6436,6437,6438,6443,6447,6448,6449],{"id":6113,"depth":490,"text":6114},{"id":6124,"depth":490,"text":6125},{"id":6148,"depth":490,"text":6149},{"id":6176,"depth":490,"text":6177,"children":6439},[6440,6441,6442],{"id":6180,"depth":944,"text":6181},{"id":6187,"depth":944,"text":6188},{"id":6220,"depth":944,"text":6221},{"id":6227,"depth":490,"text":6228,"children":6444},[6445,6446],{"id":6234,"depth":944,"text":6235},{"id":6277,"depth":944,"text":6278},{"id":6322,"depth":490,"text":6323},{"id":6364,"depth":490,"text":6365},{"id":6416,"depth":490,"text":6417},"2026-02-02","Learn what BYOK (Bring Your Own Key) means for AI hosting, how it saves you money, and why ClawNest uses this model for OpenClaw hosting.",{},"/blog/byok-ai-hosting-explained",{"title":6108,"description":6451},"blog/byok-ai-hosting-explained",[6457,6458,6459],"BYOK","Pricing","AI Hosting","dqDi35MffetA1e0fBmBLdI_8Q6gMgspIS3ZcSsxb0MI",1781298816664]