daijinma/openclaw-mcp-bridge
龙虾连接本地AI CLI,比如opencode、claude、codex等
Platform-specific configuration:
{
"mcpServers": {
"openclaw-mcp-bridge": {
"command": "npx",
"args": [
"-y",
"openclaw-mcp-bridge"
]
}
}
}Add the config above to .claude/settings.json under the mcpServers key.
---
<a id="english"></a>
MCP Server bridging local AI tools (opencode / Claude / Codex) to OpenClaw agents.
Local AI (opencode / Claude / Codex)
↓ MCP stdio (local) or HTTP (serve mode)
openclaw-mcp-bridge (per-instance process)
↓ openclaw agent CLI (local) or WebSocket (remote)
OpenClaw
↓
Agent (main / geo / ...)git clone git@github.com:daijinma/openclaw-mcp-bridge.git
cd openclaw-mcp-bridge
npm installnpx openclaw-mcp-bridge install localclawThis adds two entries to ~/.config/opencode/opencode.json:
openclaw-localclaw — runs the bridge process@localclaw — subagent you can @ in opencode TUI@localclaw 最近做了哪些事情?
@localclaw 帮我看看 geo agent 的配置Connects via openclaw agent CLI subprocess. Requires openclaw installed locally.
npx openclaw-mcp-bridge install localclaw
npx openclaw-mcp-bridge install localclaw --bin /usr/local/bin/openclawConnects via WebSocket to a remote OpenClaw Gateway server. No local openclaw CLI needed.
npx openclaw-mcp-bridge install kupuclaw --host wss://your-server:18789 --token your-tokenTo use remote mode, you need an OpenClaw server with the Gateway accessible over the network:
1. Install OpenClaw on the server
npm install -g openclaw
openclaw setup2. Configure Gateway for remote access
Edit ~/.openclaw/openclaw.json on the server:
{
"gateway": {
"port": 18789,
"bind": "lan",
"auth": {
"mode": "token",
"token": "your-secret-token"
}
}
}| Field | Value | Description | |---|---|---| | bind | "lan" or "custom" | "loopback" = localhost only, "lan" = LAN accessible | | auth.mode
Loading reviews...