MCP server (AI integration)

ManicTime includes a built-in MCP (Model Context Protocol) server. This allows AI tools that support MCP to read your recorded activity data and write tags back to ManicTime.

Requirements

ManicTime v2026.1 or newer.

How it works

The MCP server runs locally on your computer and communicates with your AI tool over stdio. On Mac and Linux, start it by running the ManicTime executable with the mcp argument. On Windows, run ManicTimeMcp.exe with no arguments. No separate installation or configuration in ManicTime is needed.

Important: MCP startup is different on Windows. Use ManicTimeMcp.exe directly on Windows. Use ManicTime mcp on Mac and Linux.

ManicTime does not route any activity data through ManicTime servers for this feature. The data you query is sent directly from your computer to the AI provider you have configured.

Executable paths

OS Default path
Windows C:\Program Files\ManicTime\ManicTimeMcp.exe
Mac /Applications/ManicTime.app/Contents/MacOS/ManicTime
Linux /opt/manictime/ManicTime

If ManicTime is installed to a different location, adjust the path accordingly.

Setting up with AI tools

All integrations use the same principle: configure your AI tool to start the ManicTime MCP server over stdio. On Mac and Linux, run the ManicTime executable with mcp as the argument. On Windows, run ManicTimeMcp.exe with no arguments. The command examples below use the Windows path by default.

Claude Code

Windows:

claude mcp add manictime-client --transport stdio -- \
  "C:\Program Files\ManicTime\ManicTimeMcp.exe"

Mac:

claude mcp add manictime-client --transport stdio -- \
  "/Applications/ManicTime.app/Contents/MacOS/ManicTime" mcp

Linux:

claude mcp add manictime-client --transport stdio -- \
  /opt/manictime/ManicTime mcp

Claude Desktop

Open Settings → Developer → Edit Config and add:

Windows:

{
  "mcpServers": {
    "manictime-client": {
      "command": "C:\\Program Files\\ManicTime\\ManicTimeMcp.exe"
    }
  }
}

Mac:

{
  "mcpServers": {
    "manictime-client": {
      "command": "/Applications/ManicTime.app/Contents/MacOS/ManicTime",
      "args": ["mcp"]
    }
  }
}

Linux:

{
  "mcpServers": {
    "manictime-client": {
      "command": "/opt/manictime/ManicTime",
      "args": ["mcp"]
    }
  }
}

Restart Claude Desktop after saving.

OpenAI Codex

Windows:

codex mcp add manictime-client -- \
  "C:\Program Files\ManicTime\ManicTimeMcp.exe"

Mac:

codex mcp add manictime-client -- \
  /Applications/ManicTime.app/Contents/MacOS/ManicTime mcp

Linux:

codex mcp add manictime-client -- \
  /opt/manictime/ManicTime mcp

Visual Studio Code (GitHub Copilot)

Option A — UI:

  1. Ctrl+Shift+P / Cmd+Shift+PMCP: Add Server

  2. Select Command (stdio)

  3. Enter the command for your platform, for example:

    Windows:

     "C:\Program Files\ManicTime\ManicTimeMcp.exe"
    

    Mac:

     "/Applications/ManicTime.app/Contents/MacOS/ManicTime" mcp
    
  4. Enter ManicTime as the server name

Option B — config file:

Create .vscode/mcp.json in your workspace:

{
  "servers": {
    "ManicTime": {
      "type": "stdio",
      "command": "C:\\Program Files\\ManicTime\\ManicTimeMcp.exe"
    }
  }
}

Open Copilot Chat and switch to Agent mode to use MCP tools.

What the AI can access

Once connected, the AI can query ManicTime activity for any time range:

  • Computer usage (active/away periods)
  • Applications used
  • Documents opened
  • Websites visited
  • Tags

The AI can also write tags back to ManicTime, allowing it to create structured time entries from raw activity data.

Privacy

The MCP server runs locally on your computer. ManicTime does not send your activity data through ManicTime servers as part of this feature.

The activity data you ask the AI to analyze is sent to the AI provider you have configured (for example Anthropic, OpenAI). Whether that provider processes data locally or in the cloud depends on the tool and its settings.