🖥️ fastmcp-linux-desktop

High-performance FastMCP Server empowering AI Agents (Claude Desktop, Cursor, Antigravity, OpenRig) with native Linux desktop and compositor control.
Provides deep primitives for Hyprland/Wayland window management, PipeWire audio routing, systemd health diagnosis and self-healing, notifications, and long-term agent memory.

⚡ Features & MCP Tool Catalog

🪟 Wayland & Hyprland Window Management

• get_windows(): Enumerate all active windows, classes, floating status, and workspace geometry.

• focus_window(target): Focus window by title, class, or hex address (e.g. Alacritty, Chrome).

• dispatch_hyprland(dispatcher, args): Run any arbitrary hyprctl dispatch instruction.

• switch_workspace(workspace_id): Switch workspace dynamically.

• get_monitors(): List connected outputs, resolutions, scale factors, and refresh rates.

🔊 Audio & DSP Routing

• get_audio_status(): Query default sink volume, mute state, and driver.

• set_volume(percentage): Smoothly adjust sink volume (0-150%).

• toggle_audio_sink(): Cycle audio output between HDMI, line-out, headphones, or Bluetooth.

🩺 System Diagnostics & Self-Healing

• get_system_health(): Instant diagnosis of failed system/user systemd units, load, and RAM.

• heal_system(): Automatically reset-failed and restart degraded units, recovering audio if stalled.

• get_resource_usage(): View filesystem disk usage and top CPU/memory consuming processes.

• desktop_notify(title, message, urgency): Trigger native desktop notifications via notify-send.

• take_screenshot(output_path): Capture Wayland screenshots using grim.

🧠 Agentic Memory & Context (Optional)

• condense_context(raw_text, query): Real-time BM25 context compression (up to 98% token reduction) via context-condenser.

• memory_retain(fact, category, importance): Long-term episodic memory storage via agent-hindsight.

• memory_recall(query, category, limit): Fast memory retrieval with recency decay.

Related MCP server: Linux System MCP Server

🚀 Installation & Usage

1. Install via pip

pip install fastmcp-linux-desktop

2. Connect to Claude Desktop / Cursor / Antigravity

Add to your claude_desktop_config.json or MCP settings:

{
  "mcpServers": {
    "linux-desktop": {
      "command": "fastmcp-linux-desktop",
      "args": ["--transport", "stdio"]
    }
  }
}

3. Run as a Persistent Daemon (SSE / HTTP)

# Run over SSE on port 8991
fastmcp-linux-desktop --sse --port 8991

# Or run as a systemd user service
systemctl --user enable --now fastmcp-linux-desktop.service

🔧 Systemd User Unit Example

Create ~/.config/systemd/user/fastmcp-linux-desktop.service:

[Unit]
Description=FastMCP Linux Desktop MCP Server
After=graphical-session.target

[Service]
Type=simple
ExecStart=/usr/local/bin/fastmcp-linux-desktop --sse --port 8991
Restart=on-failure
RestartSec=3

[Install]
WantedBy=default.target

🧪 Testing

python3 -m unittest discover tests

📄 License

MIT License. Designed and maintained by apravint.