How your AI connects

atbridge.ai plugs the AI of your choice — Claude, Cursor, a local model, any agent — straight into your inbox: every email and contact, from any provider, through Thunderbird. There are two ways to connect it, and which one you need comes down to a single question: can your assistant run terminal commands on its own?

The two ways

1. Directly, through the command line

If your assistant can run shell commands — Claude Code, Cursor’s agent, Aider, or any agent with terminal access — it just runs the atbridge command itself. Once atbridge.ai is installed, there’s nothing else to set up: no config file, no connection step.

2. Through an MCP connection

If your assistant connects over MCP (Model Context Protocol) — Claude Desktop, Cursor, VS Code, JetBrains IDEs, Antigravity, or any MCP client — you add atbridge.ai to its MCP config once. atbridge can write that config for you: pick your client in @bridge Settings → Connect your AI (MCP), or run atbridge mcp setup --client <name>. The setup pages below also walk you through it by hand.

Not sure which you have?

If your AI can open a terminal or run shell commands, use method 1 — skip MCP entirely. If it’s a chat app with no terminal, use method 2.

Set up your app

Claude Desktop No terminal — connect over MCP.

Cursor Connect over MCP, the same way as Claude Desktop.

JetBrains IDEs IntelliJ, PyCharm, WebStorm, GoLand, Rider… — one config for all.

Antigravity Google's agentic IDE + CLI, over MCP.

VS Code & other MCP clients

VS Code works the same way — one click in @bridge Settings → Connect your AI (MCP), or atbridge mcp setup --client vscode from a terminal (add --allow-writes to let the AI send / draft / edit mail; read-only otherwise). Any other MCP app uses the same atbridge mcp command. Any agent with a terminal can run atbridge directly — head straight to your first AI-written email.

Where to go next

• Try a recipe — pre-built workflows