OpenClaw Configuration File: Every Setting Explained

The openclaw.json configuration file is the control center for your OpenClaw instance. Located at ~/.openclaw/openclaw.json, it controls everything from LLM provider settings to channel configuration, security policies, logging, and skill management. This reference explains every major section and setting.

Top-Level Structure

{
  "llm": { ... },
  "channels": { ... },
  "skills": { ... },
  "memory": { ... },
  "security": { ... },
  "web_interface": { ... },
  "logging": { ... },
  "scheduler": { ... }
}

LLM Section

{
  "llm": {
    "provider": "anthropic",
    "api_key": "sk-ant-...",
    "model": "claude-3-5-sonnet-20261022",
    "max_tokens": 8192,
    "temperature": 0.7,
    "context_window": 200000,
    "context_truncation": {
      "enabled": true,
      "max_tokens": 50000,
      "strategy": "summarize_oldest"
    },
    "system_prompt": "You are a helpful personal assistant. Be concise and accurate."
  }
}

temperature: Controls randomness (0 = deterministic, 1 = creative). Use 0.3–0.5 for factual tasks, 0.7–0.9 for creative work.

system_prompt: Your agent's core personality and instructions. Changes here affect all conversations.

Channels Section

See individual channel setup guides for detailed configuration. The channels section supports: telegram, discord, slack, whatsapp, signal, and web (built-in web chat interface).

{
  "channels": {
    "telegram": {
      "enabled": true,
      "token": "...",
      "allowed_users": [123456789],
      "parse_mode": "MarkdownV2"
    }
  }
}

Memory Section

{
  "memory": {
    "enabled": true,
    "backend": "sqlite",
    "path": "~/.openclaw/memory.db",
    "vector_search": true,
    "max_memories": 10000,
    "auto_summarize": true,
    "summarize_after_messages": 100
  }
}

vector_search: Enables semantic memory search — finding relevant memories based on meaning, not just keywords. Requires the memory skill from ClawHub.

Security Section

{
  "security": {
    "allowed_origins": ["https://yourdomain.com"],
    "canvas_host": {
      "enabled": true,
      "sandbox_level": "strict"
    },
    "tools": {
      "allow": ["read_file", "write_file", "web_search"],
      "deny": ["execute_command"]
    }
  }
}

Scheduler Section

{
  "scheduler": {
    "enabled": true,
    "timezone": "America/New_York",
    "tasks": [
      {
        "name": "morning_briefing",
        "cron": "0 8 * * *",
        "action": "run_skill",
        "skill": "daily-briefing",
        "output_channel": "telegram"
      }
    ]
  }
}

Web Interface Section

{
  "web_interface": {
    "enabled": true,
    "port": 8080,
    "host": "127.0.0.1",
    "auth": {
      "enabled": true,
      "type": "totp",
      "secret": "YOUR_TOTP_SECRET"
    }
  }
}

Always bind to 127.0.0.1 (not 0.0.0.0) and use nginx as a reverse proxy.

Frequently Asked Questions

Where can I find a complete openclaw.json schema?

The full JSON Schema is at openclaw/schemas/config.schema.json in the OpenClaw repository. Use it with VS Code for auto-complete and validation when editing your config.

Does OpenClaw hot-reload the config file?

Most settings are loaded at startup. Use python -m openclaw reload to apply config changes without a full restart. Some settings (LLM provider, port) require a restart.

Can I split configuration across multiple files?

Yes. Use "include": "path/to/partial-config.json" to include additional config files. Useful for keeping sensitive credentials in a separate file with restricted permissions.