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

# Connect Your Agent

> Use Monk with VS Code, Claude, Gemini, Codex, Cursor, and other coding agents

When your coding agent can talk to Monk, the full loop closes. Your agent writes and changes the code, Monk deploys it, provisions infrastructure, and keeps everything running. You describe what you want, walk away, and come back to working software on your cloud — no architectural constraints, no manual DevOps steps in between.

## Monk MCP Is Enabled Automatically

The Monk MCP server is enabled automatically as part of the [one-time setup](/getting-started/installation). No command palette step, no manual toggle — it just works.

If you ever need to check the status, the setup screen in the Monk chat window shows whether the MCP server is running. You can get back to it from the Monk icon in the bottom-left of your IDE's status bar.

<Note>
  Claude Code, Codex, and Cursor can also install Monk directly through the [Monk agent plugin](/getting-started/agent-plugin-installation). Use that flow if you do not want to install the IDE extension first.
</Note>

## Connect Your Agent

The MCP server is running, but each agent needs one more step on its side to trust it. Follow the guide for your agent:

<CardGroup cols={2}>
  <Card title="VS Code / GitHub Copilot" icon="code" href="/getting-started/mcp-vscode-copilot">
    Enable Monk tools in Copilot chat
  </Card>

  <Card title="Cursor" icon="arrow-pointer" href="/getting-started/mcp-cursor">
    Approve Monk tool usage in Cursor
  </Card>

  <Card title="Windsurf" icon="wind" href="/getting-started/mcp-windsurf">
    Enable MCP tools in Cascade
  </Card>

  <Card title="Claude Code" icon="sparkles" href="/getting-started/mcp-claude-code">
    Project-local MCP config for Claude Code
  </Card>

  <Card title="Codex" icon="terminal" href="/getting-started/mcp-codex">
    Codex CLI or the Codex IDE extension
  </Card>

  <Card title="Gemini CLI" icon="message" href="/getting-started/mcp-gemini-cli">
    Project-scoped settings for Gemini
  </Card>

  <Card title="Antigravity" icon="wand-magic-sparkles" href="/getting-started/mcp-antigravity">
    MCP management UI in Antigravity
  </Card>
</CardGroup>

## Two Patterns

**IDE-embedded agents** like Copilot, Cursor, Windsurf, and Antigravity run inside the same IDE as Monk. Monk sets up MCP automatically — you just need to allow the agent to use Monk tools.

**Standalone agents** like Claude Code, Gemini CLI, and Codex run in a separate terminal. Keep Monk open in your IDE with the project loaded, then launch the standalone agent in the same directory. Monk writes the config file the agent needs automatically.

## What Monk Configures

During setup, Monk writes client-native config for the current project:

| Agent                | Config location                        |
| -------------------- | -------------------------------------- |
| VS Code / Copilot    | `.vscode/mcp.json`                     |
| Cursor / Windsurf    | `.cursor/mcp.json`                     |
| Claude Code          | `~/.claude.json` (project-local entry) |
| Gemini / Antigravity | `.gemini/settings.json`                |
| Codex                | `.codex/config.toml`                   |

Monk only manages its own entries. Disabling Monk MCP removes only the Monk entry from these files.

## Next Steps

<CardGroup cols={2}>
  <Card title="First deployment" icon="rocket" href="/getting-started/first-deployment">
    Deploy your first application through your agent
  </Card>

  <Card title="MCP support details" icon="plug" href="/features/mcp-support">
    How Monk MCP works under the hood
  </Card>

  <Card title="Agent plugin install" icon="plug-circle-plus" href="/getting-started/agent-plugin-installation">
    Direct plugin setup for Claude Code, Codex, and Cursor
  </Card>
</CardGroup>