OpenClaw: hardware requirements, local model setup, and context planning

What it is

Official runtime requirements

• Node runtime requirement: OpenClaw install docs require Node 24.16+ or 26.1+, and recommend Node 26. Installer provisions Node if missing. (OpenClaw install docs)
• Platform support: OpenClaw install docs list macOS, Linux, and Windows as supported installation targets. (OpenClaw install docs)
• Local-model memory note: OpenClaw local-model docs state memory depends on model weights, context, runtime, and host load; managed llama.cpp recipes use 64K context and smallest curated recipe has an 8 GiB host-memory floor. (OpenClaw local models docs)

Local model connection snippets (official docs)

openclaw onboard
# choose Ollama: Cloud + Local, Cloud only, or Local only
openclaw models list --provider ollama

models: {
  providers: {
    ollama: {
      api: 'ollama',
      baseUrl: 'http://host:11434',
    },
  },
}

models: {
  providers: {
    lmstudio: {
      api: 'openai-responses',
      baseUrl: 'http://127.0.0.1:1234/v1',
      models: [{ id: 'my-local-model', contextWindow: 196608 }],
    },
  },
}

models: {
  providers: {
    local: {
      api: 'openai-completions',
      baseUrl: 'http://127.0.0.1:8000/v1',
      apiKey: 'sk-local',
    },
  },
}

# Managed local server path from OpenClaw local-model docs
openclaw onboard
# choose Managed local server (llama.cpp plugin)

Why long context and tool-calling models matter

• OpenClaw local-model docs emphasize that a model passing short prompts may still fail full agent turns because tool schemas, history, and prompts consume additional context.
• OpenClaw setup verifies a real tool call before switching the default local model in managed setup flow.

Prefilled calculator

The fit matrix below is an LLMRAM estimate at Q4_K_M and batch=1 for planning. It is not an official vendor minimum requirement.

Local model fit table by memory tier (LLMRAM estimate)

The fit matrix below is an LLMRAM estimate at Q4_K_M and batch=1 for planning. It is not an official vendor minimum requirement.

FAQ

Verification notes

• The runtime requirements section only includes statements present in official framework docs listed on this page.
• Fit tables are LLMRAM planning estimates derived from current model pages and calculator assumptions.
• When a runtime-specific snippet is missing in provided official links (for example Hermes llama.cpp), this page marks the gap explicitly.

Official sources used

• OpenClaw GitHub
• OpenClaw install docs
• OpenClaw Ollama provider docs
• OpenClaw Ollama setup docs
• OpenClaw local models docs

Internal links

• LLMRAM calculator home
• Agent frameworks index
• Mistral 7B Instruct v0.3
• Qwen 3.8 27B
• GLM 5.3 (Flash)
• RTX 4090 page
• Apple M3 Max 128GB page