> ## Documentation Index
> Fetch the complete documentation index at: https://docs.ultrametric.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Connect MCP

> Add Ultrametric tools to an agent that supports remote MCP and OAuth.

In your agent host's remote MCP settings, add a connection with this server URL:

```text theme={null}
https://api.ultrametric.ai/mcp
```

Complete the OAuth sign-in offered by the host. Choose the Ultrametric account and organization you want to use. Keep tokens and sign-in codes out of shared messages and repository files.

The host must support remote MCP and OAuth. Use its setup instructions to find the connection settings; the names and locations differ between applications.

## Check the connection

Ask your agent:

```text theme={null}
Use Ultrametric to check my connection and selected organization.
Then list the available processes and read one relevant to this project.
```

The agent should call `status`, then `list_processes`, then `get_process` with a returned ID. Check the account and environment in the status result. A successful guide response includes its instructions and result schema.

If sign-in fails or the catalog is empty, see [troubleshooting](/troubleshooting). For the exact tool inputs, see [MCP tools](/mcp/tools).

## Continue across sessions

Use [saved context](/guides/resume-work) to find earlier company records and runs. CLI and MCP can access the same hosted records when signed in to the same organization. Their login sessions are separate.

Tools return text and structured data that your agent can use. Some hosts can also show a process view. If a view is unavailable, ask the agent to summarize the returned data.

For staging work, use a separately named connection to `https://api.ultrametric.dev/mcp` and sign in for that environment. Keep production and staging connections distinct.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.