Set up MCP and more!
Connecting your Rill project to your AI Agent has never been easier!
Prerequisites
Before you begin, make sure you have:
- A MCP Client (Claude Desktop, ChatGPT Desktop, etc.)
- Access to a Rill Project
For detailed steps, please refer to our documentation on MCP Servers
Step 1: Access your Rill Project's AI tab.

In your project's AI tab, you'll be able to create a [user based token] and copy the MCP config.json directly in the UI. Simple paste this into your application of choice.
Step 2: Add config.json into Client
Depending on your client application, paste the config from step 1.
- Anthropic's Claude Desktop Documentation
- OpenAI's ChatGPT Documentation
- Cursor Documetnation
- Windsurf Documentation
Step 3: Start Querying your Agent about your Rill metrics.
A below is an example chat with Claude using our MCP server to get information on the commit history for one of our demo projects.

Step 4: Agent Customization
If you watched the whole clip above, you'll see that Claude tries to make code a webapp to visualize the information that it found, but you'll see it takes time. (fast-forwarded in the clip) Instead, what you can do is provide the Agent context and instructions on how you'd like the Agent to behave.
I highly recommend taking a look at our series of videos on Conversation BI with Claude and Rill.