Skip to main content
Empower your AI agents by connecting Ruba as a Model Context Protocol (MCP) server.

What exactly is MCP?

MCP is a protocol for wiring tools into AI agents. It can substantially boost what your AI agents are capable of by feeding them real-time data and context. Ruba hosts a remote MCP server that you can plug into from most AI clients.

How it operates

You need an MCP-aware agent environment to use Ruba through MCP. A couple of well-known ones are Claude Desktop and Cursor.

Linking up with Ruba MCP

Ruba provides two MCP servers:
  • Production: https://mcp.getruba.com/mcp/ruba - Hook into your live Ruba organization
  • Sandbox: https://mcp.getruba.com/mcp/ruba-sandbox - Hook into the Ruba sandbox environment for testing
When you can enter a MCP URL, pick one of the URLs above based on your environment. If you must specify a command, use:
For sandbox:

Cursor

Inside .cursor/mcp.json, add:
For sandbox:

Windsurf

Inside mcp_config.json, add:
For sandbox:

Codex

Append the following to your ~/.codex/config.toml:
Then run:
For sandbox:
Then run:

Claude Code

Run the following command:
For sandbox:

OpenCode

Inside opencode.jsonc, add:
For sandbox:

ChatGPT

MCP is currently available only to paid users in beta on ChatGPT web, by turning on Developer Mode. Once Developer Mode is active, head to SettingsConnectors and register the MCP server using https://mcp.getruba.com/mcp/ruba. For sandbox, use https://mcp.getruba.com/mcp/ruba-sandbox instead.

Claude Desktop

Head to SettingsConnectors and click Add custom connector. Name it “Ruba” and supply https://mcp.getruba.com/mcp/ruba as the server URL. For sandbox, use https://mcp.getruba.com/mcp/ruba-sandbox as the server URL instead. Save, then click Connect to link up with your Ruba organization.