Model Context Protocol (MCP)

The Model Context Protocol (MCP) is an open standard that lets AI agents interact with external tools and services. Think of MCP servers as plugins that give your agent new abilities — from reading files and running code to managing Linear issues and sending Slack messages.

What is MCP?

MCP provides a standardized way for AI models to:

• Call tools — Execute actions in external services
• Read resources — Access data from connected systems
• Use prompts — Load specialized instructions and templates

Each MCP server exposes a set of tools that the agent can use during task execution. The agent decides which tools to call based on your request.

Built-in MCP Servers

The desktop app comes with several pre-configured MCP servers:

Sandbox Server

Execute code safely during agent tasks:

The sandbox uses OS-level isolation (macOS Seatbelt / Linux Bubblewrap) to prevent unauthorized access.

Linear Server

Full issue management with 18 tools:

Media Generation Server

Create images and videos with AI:

Memory Server

Long-term memory that persists across sessions:

Google Services Server

Access 79 Google Workspace tools (scope-filtered per your granted permissions):

• Gmail — Search, read, send, reply, manage labels
• Calendar — Create, update, list events
• Drive — Search, read, create files and folders
• Sheets — Read and write spreadsheet data
• Docs — Create and edit documents
• Meet — Schedule and manage meetings
• Tasks — Create and manage task lists
• Contacts — Search and manage contacts

Speech Server

Voice input and output:

Install MCP Servers

From the Presets Gallery

1. Open Settings > MCP in the desktop app
2. Browse the presets gallery
3. Click Install on any preset
4. The server is automatically configured and available to your agent

Remote OAuth Servers

Some servers (Notion, Figma, Granola) use OAuth for authentication:

1. Click Install on a Remote OAuth preset
2. A browser window opens for authorization
3. Grant the requested permissions
4. The token is securely stored and the server is ready to use

Manual Configuration

Add custom MCP servers by editing the configuration file:

Application config — ~/.<app-slug>/mcp.json:

{
  "mcpServers": {
    "my-server": {
      "command": "npx",
      "args": ["-y", "@my-org/mcp-server"],
      "env": {
        "API_KEY": "your-api-key"
      }
    }
  }
}

Claude Code shared config — ~/.claude/settings.json:

{
  "mcpServers": {
    "shared-server": {
      "command": "node",
      "args": ["path/to/server.js"]
    }
  }
}

Both configuration sources are merged at startup.

Supported Transports

Tips

• Start with presets — The gallery covers the most common use cases
• Check permissions — OAuth servers only access the scopes you grant
• Restart after changes — MCP configuration is loaded at app startup
• View available tools — After connecting, browse the tools list in Settings > MCP to see what's available