CLAWME

How to connect an Agent to Claw Me

Connect Codex, Claude Code, Cursor, OpenClaw, or another compatible client, then choose exactly what it can access.

Connect your assistant

Try this with your assistant

Example prompt. Your assistant needs the relevant Claw Me access to complete account work.

From conversation to workspace

  1. Choose your client

    Open Plugins and use the setup instructions for your client. Keep API keys and authorization codes out of public chats.

  2. Review the request

    Open Settings → Connected Agents. Check the Agent name and exact access before approving. Entering a code only opens the request for review.

  3. Verify and reuse

    Ask the Agent to perform a permitted read. Reuse its connection until it expires or you revoke it.

Check the result

Review the result in your own account. A successful connection or copied prompt does not mean the requested action has completed.

Choose a client →

What to know before you start

MCP/API clients can connect on every plan. A2A connections require Basic or Plus and a compatible Agent endpoint. Connection permission is separate from account feature settings.

See current plans and storage allowances and how Claw Me handles your data.

Common questions

Why does a connected Agent still get access denied?

Its scopes, account feature settings, plan, and resource access must all permit the operation. Review the missing permission instead of repeatedly reconnecting.

How do I disconnect it?

Revoke the connection in Settings → Connected Agents. This stops future credential access; it cannot recall information already read.