Weavers / MCP documentation

Connect an assistant

Set up a Weavers MCP connection, sign in, and try your first request.

Before you start

Use a Weavers account with a verified email address.
Use an assistant enabled for your Weavers deployment. The client must support remote Streamable HTTP and the configured OAuth sign-in flow.
Client registration and access are managed by Weavers. Adding a URL to an arbitrary MCP client does not register that client or grant access.

In your assistant's connection settings, add the Weavers connection supplied for your account. For a custom remote connection, use the server URL below and the client configuration supplied by Weavers.
Choose HTTP / Streamable HTTP if the application asks for a transport. Menu names depend on the assistant.
If your assistant is not available in AI connections, contact Weavers before attempting a custom setup.

1. Add the Weavers connection

https://www.weavers.app/api/mcp

Start authorization from your assistant. Follow the browser flow and sign in to your existing Weavers account.
Check which account and permissions are requested. Reading source code, previewing data, preparing drafts, and proposing code each require their own permission.
If your deployment asks you to link the connection first, open Weavers → AI connections, select Connect, and finish the confirmation in the same browser session.
Return to the assistant after authorization. You do not need to paste a Weavers password, browser cookie, or service token into a prompt.

Ask the assistant to list your five most recently updated notebooks.
It should use list_my_notebooks and return titles from your account. The default Research Notebook can appear even if you have not saved it yet.
If the assistant asks for sign-in or an additional permission, complete that flow before retrying.

3. Try a small first request

Show me my five most recently updated Weavers notebooks.

Find a strategy project, then inspect its revisions and dataset versions.
Read source code or preview a dataset only when you want to share that information with the connected assistant.
For a draft or code proposal, open the returned Weavers link and review it there.
Manage or disconnect access at /settings/ai-connections. Disconnect blocks subsequent tool calls; it does not erase information already shared with the assistant.