Get Started
Integrating the Pieces MCP with OpenClaw brings your workflow context directly into this open-source continuous AI agent runtime. OpenClaw (formerly ClawdBot, formerly Moltbot) runs as a persistent Node.js service—unlike chatbots that respond to one-off prompts, it runs 24/7, executing tasks proactively via cron jobs and event listeners.
With Pieces MCP connected, OpenClaw gains access to your Long-Term Memory. It can autonomously query your past work, generate standups, monitor recent activity, and surface relevant context without you asking.
Prerequisites
Installing PiecesOS & Configuring Permissions
Follow the instructions below for a detailed guide on setting up and configuring PiecesOS to correctly pass captured workflow context to the Pieces MCP server.
Setting Up OpenClaw
OpenClaw executes MCP tools via MCPorter, its built-in MCP management layer. Edit ~/.openclaw/workspace/config/mcporter.json.
The official Pieces MCP skill for OpenClaw documents MCP-only integration: point mcp-remote at the /mcp endpoint (/model_context_protocol/2025-03-26/mcp), not the legacy /sse path. That matches how OpenClaw and MCPorter expect to bridge Pieces.
Install mcp-remote globally with a pinned version for security:
npm install -g mcp-remote@0.1.38
See MCP Bridge for why we recommend a locally installed binary over npx.
Local setup (MCP with mcp-remote — recommended)
When OpenClaw and PiecesOS run on the same machine, use the localhost MCP URL with the mcp-remote bridge:
{
"mcpServers": {
"pieces": {
"command": "mcp-remote",
"args": [
"http://localhost:39300/model_context_protocol/2025-03-26/mcp"
]
}
}
}
Remote setup (ngrok or other HTTPS tunnel)
When OpenClaw runs on a different machine than PiecesOS, expose PiecesOS (port 39300) with a tunnel and use the same MCP path on your tunnel base URL:
{
"mcpServers": {
"pieces": {
"command": "mcp-remote",
"args": [
"https://YOUR_NGROK_URL.ngrok-free.app/model_context_protocol/2025-03-26/mcp"
]
}
}
}
See Tunneling for tunnel setup. For step-by-step tunnel checks, curl validation, and gateway restarts, follow the ClawHub skill.
Example Use Cases
Once Pieces MCP is connected to OpenClaw, you can automate workflows like:
Autonomous daily standup: Schedule OpenClaw to run every morning, query yesterday's workstream summaries, and post a formatted standup to your Slack or Teams channel.
Meeting prep: Before a calendar event, OpenClaw searches audio transcriptions and workstream summaries for context related to the meeting topic and drafts a brief for you.
Automated debugging log: When OpenClaw detects a production alert, it queries recent workstream events for error-related content and creates a pieces_memory entry with the incident context.
Verification
Security Note
OpenClaw can run with permissionMode: 'bypassPermissions' to execute tools autonomously. When combined with Pieces MCP write tools (like create_pieces_memory), this is powerful but should be used carefully. Consider:
- Running OpenClaw in Docker with limited filesystem access
- Disabling write tools in MCPorter if running fully autonomously
- Monitoring execution logs
Updating
Edit ~/.openclaw/workspace/config/mcporter.json, update the URL, then restart the OpenClaw gateway (for example openclaw gateway restart from ~/.openclaw/workspace, as in the official skill).
Troubleshooting
If you're experiencing issues integrating Pieces MCP with OpenClaw:
You're now set to enhance your OpenClaw workflow with powerful context retrieval through Pieces MCP. Happy coding!