MCP Server

What you need

• Mach Triage desktop app installed and running.
• The MCP server enabled in Settings → MCP Server.
• An MCP-compatible AI tool (Claude Code, Claude Desktop, Cursor, Windsurf, etc.) configured to run the bundled binary.

Step 1: Enable the MCP server

Open the Mach Triage desktop app, go to Settings → MCP Server, and flip the Enable toggle. The server does not run a persistent port—your AI tool launches it as a subprocess on demand via stdin/stdout.

When disabled, the binary still starts but immediately rejects all requests so your AI tool gets a clear error rather than stale data.

Step 2: Find the binary path

The MCP binary (mach-triage-mcp) is bundled inside the app installation alongside the main executable.

• macOS: /Applications/Mach Triage.app/Contents/MacOS/mach-triage-mcp
• Windows: C:\Program Files\Mach Triage\mach-triage-mcp.exe
• Linux: same directory as the mach-triage main binary

The binary auto-detects your database from the platform app-data directory. You can override this with the --db-path flag or the MACH_TRIAGE_DB_PATH environment variable if your app data is in a non-standard location.

Step 3: Configure your AI tool

Claude Code

Add the server to your project or global MCP config. The easiest way is to run claude mcp add from your terminal:

claude mcp add mach-triage /Applications/Mach\ Triage.app/Contents/MacOS/mach-triage-mcp

Or add it manually to your MCP settings JSON:

{
  "mcpServers": {
    "mach-triage": {
      "command": "/Applications/Mach Triage.app/Contents/MacOS/mach-triage-mcp"
    }
  }
}

Claude Desktop

Edit ~/Library/Application Support/Claude/claude_desktop_config.json (macOS) or %APPDATA%\Claude\claude_desktop_config.json (Windows) and add the same mcpServers block above.

Cursor / Windsurf

Open your MCP settings (usually in preferences under "Model Context Protocol" or "AI Tools") and add a new server pointing to the binary path. Use the same JSON structure shown above.

Available tools

Once connected, your AI tool will see the following tools:

Read tools (Free and Pro)

• list_workspaces — list all active issue-tracking workspaces.
• get_active_workspace — get details for the currently active workspace.
• list_tickets — list or filter tickets, with optional workspace_id and status filters.
• search_tickets — full-text search across ticket titles, descriptions, and external keys (e.g., PROJ-42).
• get_ticket — full detail for a ticket by database ID or external key.
• list_work_logs — list work logs for a ticket or recent logs globally.

Write tools (Free and Pro)

• add_work_log — create a new work log on a ticket. Jira logs queue for sync to Atlassian; Linear and GitHub track locally.
• add_comment — add a comment to a ticket, queued for provider sync.

Pro only

• search_armory — full-text search across your Mach Armory notes and snippets.

Plan limits

The MCP server respects the same workspace limits as the desktop app. On Free, tools that query by workspace are restricted to your first workspace slot. On Pro, all workspaces are accessible and search_armory is available.

Data and privacy

The MCP server reads directly from your local SQLite database—the same file the desktop app uses. No data is sent to Mach Triage cloud services. Your provider credentials (Jira OAuth, Linear API keys, GitHub tokens) are not exposed through MCP. Write operations (work logs, comments) insert locally first and queue for sync using the same sync engine as the desktop app.

Troubleshooting

• "MCP server is disabled in Mach Triage settings." — flip the Enable toggle in Settings → MCP Server.
• Binary not found — verify the app is installed and the path matches your platform. Check the Step 2 paths above.
• Empty workspace list or no tickets — confirm your active workspace has synced. Try Sync Now from the desktop triage palette.
• search_armory returns a Pro-required error — Mach Armory is a Pro feature. Upgrade to access it.
• Wrong database path — use --db-path /path/to/mach-triage.db or set MACH_TRIAGE_DB_PATH in the server's environment.