Set Up Your AI Agent

Why this step matters

You've built a rich data context — table schemas, column descriptions, quality checks. But that context is sitting in files on your machine. Your AI coding tool doesn't know it exists yet.

MCP (Model Context Protocol) is the bridge. It's a lightweight protocol that lets AI tools call external commands — in this case, Bruin CLI. Once you configure MCP, your AI agent can:

Read your asset files to understand table structures and business context
Run bruin query to execute SQL against your warehouse
Access Bruin documentation to know how to use the CLI correctly

Without MCP, you'd have to manually copy-paste schema info into prompts. With it, the agent has direct, structured access to everything.

What you'll do

Add Bruin as an MCP server in your AI coding tool. Pick the tab that matches the tool you're using.

Cursor

Open Cursor Settings (Cmd/Ctrl + ,)
Navigate to MCP & Integrations
Click Add Custom MCP
Add the following configuration:

{
  "mcpServers": {
    "bruin": {
      "command": "bruin",
      "args": ["mcp"]
    }
  }
}

Restart Cursor

To verify it's working, open the AI chat and ask: "What connections are available in this Bruin project?" — the agent should list the connection you configured in Step 2.

Claude Code

Run this single command in your terminal:

claude mcp add bruin -- bruin mcp

That's it. Claude Code will now use Bruin MCP for any Bruin-related operations.

To verify it's working, ask Claude Code: "What connections are available in this Bruin project?" — it should list the connection you configured in Step 2.

Codex

Add the following to your ~/.codex/config.toml file:

[mcp_servers.bruin]
command = "bruin"
args = ["mcp"]

Restart Codex for the changes to take effect.

To verify it's working, ask Codex: "What connections are available in this Bruin project?" — it should list the connection you configured in Step 2.

Your AI coding tool now has a live connection to Bruin CLI. When the agent needs to understand your data or run a query, it calls Bruin through MCP rather than guessing. This means the agent works with your actual schema, your actual data, and your actual business context — not hallucinated table names.