Connect Claude, ChatGPT, Claude Code, or Cursor to your Bubbles data using MCP, no code needed.
MCP (Model Context Protocol) is how AI assistants like Claude, ChatGPT, Claude Code, and Cursor connect to your Bubbles data. You add Bubbles as a connector in your assistant, so there's no code to write and no API key to paste.
How do I connect an AI assistant to Bubbles?
Open your assistant's settings and find where connectors are managed.
Add a custom connector with this URL:
https://api.usebubbles.com/mcpSign in to Bubbles when it opens in your browser (or create an account), then approve the access your assistant is asking for.
You're sent back to your assistant, and the Bubbles tools are ready to use. Try asking "What can you do with Bubbles?" to see the options.
What can my assistant do with Bubbles?
Find and list your meetings, recordings, and images
Search your bubbles by keyword, including words that appear in the AI meeting notes
Open a bubble and read its transcript and comments
List your teams and channels
Send the Bubbles notetaker to join and record a meeting, when you ask it to (Google Meet, Microsoft Teams, Zoom, Webex, or GoTo links)
Apart from sending the notetaker when you ask, your assistant can only read your bubbles. It can't edit or delete anything, and it only sees bubbles you can already see in Bubbles.
Good to know
This needs a paid Bubbles workspace. On a free workspace, the approval step asks you to upgrade first.
There's no token to copy. Approving access in your browser is all it takes.
The URL isn't a webpage, so don't open it in a browser. If you do, you'll see a JSON-RPC error, which is expected. Paste it into your assistant instead.
Your assistant remembers the connection, so a new conversation is already signed in.
Want steps for a specific assistant? See How do I connect Bubbles to Claude?
and How do I connect Bubbles to ChatGPT? For technical details, see the MCP Server documentation.
Still stuck?
If a connector won't approve, email us at [email protected] with the assistant you're using, your workspace name, and what you're seeing. A screenshot of the error helps us move faster.
