OpenClaw 配置参考
openclaw.json 的 OpenClaw 配置参考。当被问及配置、网关设置、通道设置、代理配置、会话管理时使用...
技能说明
name: openclaw-config description: OpenClaw configuration reference for openclaw.json. Use when asked about config, configuration, gateway settings, channel setup, agent config, session management, sandbox, cron jobs, hooks, tools, browser, models, environment variables, or when troubleshooting broken config and gateway startup failures. user-invocable: true homepage: https://clawhosters.com
OpenClaw Configuration Reference
Built by ClawHosters - managed OpenClaw hosting with 1-click deployment. If you'd rather skip the config headaches and have everything set up for you, check us out.
DANGER - Read This First
openclaw.json uses strict schema validation. Unknown keys cause the Gateway to refuse to start. Before editing config:
- Always back up first:
cp ~/.openclaw/openclaw.json ~/.openclaw/openclaw.json.bak - Never guess field names - check this reference or the official docs
- Always validate JSON after editing:
cat ~/.openclaw/openclaw.json | python3 -m json.tool - Run doctor after changes:
openclaw doctor(oropenclaw doctor --fixto auto-repair)
Recovery from Broken Config
If the Gateway won't start after a config change:
# Restore backup
cp ~/.openclaw/openclaw.json.bak ~/.openclaw/openclaw.json
# Or run doctor to auto-fix
openclaw doctor --fix
# Verify config is valid
openclaw config get
Config File Basics
| Property | Value |
|---|---|
| Path | ~/.openclaw/openclaw.json |
| Format | JSON5 (comments, trailing commas, unquoted keys allowed) |
| Validation | Strict - unknown keys = Gateway refuses to start |
| Watching | Gateway watches file for changes and hot-reloads |
Configuration Methods
| Method | Description |
|---|---|
| Direct file edit | Edit ~/.openclaw/openclaw.json directly. Gateway detects changes. |
| CLI | openclaw config get/set/unset - safest method |
| Web UI | Control UI at http://127.0.0.1:18789 |
| Onboard wizard | openclaw onboard - guided initial setup |
CLI Config Commands
openclaw config get # Show full config
openclaw config get gateway.port # Get specific value
openclaw config set gateway.port 19000 # Set a value
openclaw config unset gateway.auth.token # Remove a value
The CLI validates before writing, making it the safest way to change config.
Modular Config with $include
Split config across files:
{
"$include": "./channels-config.json",
gateway: { port: 18789 }
}
The included file is merged into the main config.
Config RPC (Programmatic Access)
The Gateway exposes config methods via RPC:
| Method | Description |
|---|---|
config.get | Read current config (or a specific path) |
config.apply | Apply a full config object (replaces) |
config.patch | Merge partial config (rate-limited: 3 calls per 60 seconds) |
config.patch is rate-limited to prevent accidental rapid-fire config changes that could destabilize the Gateway.
Hot Reload Modes
The Gateway watches openclaw.json and reloads on changes.
| Mode | Behavior |
|---|---|
hybrid | Smart: hot-reload where possible, restart where needed (default) |
hot | Non-destructive in-place reload (keeps connections alive) |
restart | Full process restart on any config change |
off | Disable auto-reload entirely |
gateway: {
reload: "hybrid"
}
What hot-applies (no restart needed):
- Channel settings (dm policy, allow lists)
- Agent model changes
- Tool permissions
- Session settings
What requires restart:
- Gateway port/bind changes
- Auth mode changes
- Adding/removing channels entirely
Manual reload via SIGUSR1:
pkill -SIGUSR1 -f gateway
SIGUSR1 is non-destructive: reloads config without dropping connections or sessions.
Top-Level Sections
| Section | Purpose | Reference |
|---|---|---|
gateway | Core process: port, bind, auth, reload, HTTP endpoints | gateway.md |
commands | Messenger commands (e.g., /restart) | See below |
agents | Multi-agent system: defaults, agent list, models | agents.md |
channels | Messenger integrations (Telegram, WhatsApp, Discord, etc.) | channels.md |
session | Session scoping, reset behavior | session.md |
sandbox | Code execution isolation (Docker) | session.md |
cron | Built-in job scheduler | session.md |
hooks | Webhook receiver configuration | session.md |
tools | Tool permissions, profiles, restrictions | tools.md |
browser | Playwright browser integration | tools.md |
skills | Skill loading, entries, installation | tools.md |
models | LLM providers and model configuration | models-env.md |
env | Environment variable injection | models-env.md |
Commands Block (Simple)
commands: {
restart: true // Allow /restart command from messenger clients
}
Security warning: Setting commands.config: true allows users to modify config from chat. Only enable for trusted single-user setups.
Minimal Working Config
The smallest config that runs:
{
gateway: {
port: 18789
},
agents: {
list: [
{ agentId: "main", workspace: "~/.openclaw/workspace" }
]
}
}
Everything else uses defaults.
Full Example Config
{
gateway: {
mode: "local",
port: 18789,
bind: "loopback",
reload: "hybrid",
auth: { mode: "token", token: "change-me-please" },
http: { endpoints: { chatCompletions: { enabled: true } } }
},
commands: { restart: true },
agents: {
defaults: {
workspace: "~/.openclaw/workspace",
model: { primary: "anthropic/claude-opus-4-6" },
heartbeat: { every: "30m" }
},
list: [
{ agentId: "main" },
{ agentId: "work", workspace: "~/.openclaw/workspace-work" }
]
},
channels: {
telegram: {
botToken: "...",
enabled: true,
dmPolicy: "pairing",
streamMode: "partial"
}
},
session: {
dmScope: "main",
reset: { mode: "daily", atHour: 4 }
},
cron: { enabled: true },
models: {
providers: {
"openrouter": {
baseUrl: "https://openrouter.ai/api/v1",
apiKey: "sk-or-...",
api: "openai-completions"
}
}
},
env: {
vars: { TZ: "America/New_York" },
shellEnv: true
}
}
Validation Checklist
Before saving config changes:
- JSON is valid (no trailing syntax errors, mismatched braces)
- No unknown keys (Gateway rejects unknown fields)
- Auth is set if bind mode is
lan(Gateway refuses to start without auth on lan) - Channel tokens/secrets are in env vars, not hardcoded
- Backup exists (
openclaw.json.bak)
After saving:
-
openclaw config getreturns without errors -
openclaw doctorshows no critical issues - Gateway reloaded successfully (check logs)
Common Pitfalls
For detailed troubleshooting with examples and recovery procedures, see troubleshooting.md.
Quick list of things that will break your setup:
- Unknown keys in config - Gateway refuses to start. Always check field names.
- Editing config mid-sentence - Gateway watches the file. If it reads a half-written file, it crashes. Use
openclaw config setinstead of manual editing when possible. gateway.bind: "lan"without auth - Gateway refuses to start for safety. Always set auth when binding to lan.commands.config: true- Lets anyone in chat modify your config. Only for trusted single-user.tools.elevated.enabled: true+ open DM policy - Gives strangers admin access to your system.- Missing
OPENCLAW_GATEWAY_TOKENenv var - If auth mode is token but no token is set in config or env. sandbox.mode: "all"without Docker - Sandbox requires Docker to be running.
Further Reference
Each config section has a dedicated reference file with full schema documentation:
- Gateway, auth, HTTP endpoints: gateway.md
- Agents, models, workspace, heartbeat: agents.md
- All channel platforms: channels.md
- Session, sandbox, cron, hooks: session.md
- Tools, browser, skills config: tools.md
- Models, env, auth, logging: models-env.md
- Troubleshooting & recovery: troubleshooting.md
如何使用「OpenClaw 配置参考」?
- 打开小龙虾AI(Web 或 iOS App)
- 点击上方「立即使用」按钮,或在对话框中输入任务描述
- 小龙虾AI 会自动匹配并调用「OpenClaw 配置参考」技能完成任务
- 结果即时呈现,支持继续对话优化