kill-process-mcp exposes two natural-language-driven tools to manage OS processes: process_list and process_kill. You can ask your MCP-compatible AI to list running processes with filters like CPU or RAM usage, specific names, users, statuses, or to terminate selected processes. The server is designed to help you quickly identify resource hogs and safely terminate them when needed. Typical use cases include listing the top CPU or memory consumers and then issuing a command to kill a chosen process. It’s intended for cross-platform use via uv, with Python-based tooling behind the scenes, so it works on macOS, Windows, and Linux when paired with a compatible MCP client.

To interact with the server, configure your MCP client (Claude Desktop, Cursor, or other compatible client) to point at the kill-process-mcp server using the uvx-based launcher. Once configured, you can issue prompts like “List the top 5 CPU-heavy processes” or “Kill the process named Spotify” and the assistant will translate them into the appropriate tool calls under the hood. The tools are intentionally straightforward: process_list gathers and sorts processes, and process_kill terminates a selected process. Use these to diagnose system performance issues and safely remove unwanted processes without manual shell interaction.

Prerequisites

• Python 3.13 or higher
• uv (the Python async process runner)
• MCP-compatible LLM client (e.g., Claude Desktop, Cursor) to interface with the MCP server

Installation steps (preferred uvx method)

1. Install uv if it’s not already installed:

pip install uv

macOS users may also install via Homebrew

# optional alternative if you prefer Homebrew-managed uv
brew install uv

2. Run the server using uvx (preferred, no cloning required):

uvx install kill-process-mcp

Note: The above installs the latest published version of the kill-process-mcp and makes it available to your MCP client via a shim. If you prefer to run directly from source, see the alternative manual method below.

Alternative: clone the repository and run manually

1. Clone the repository and install dependencies:

git clone https://github.com/misiektoja/kill-process-mcp.git
cd kill-process-mcp
uv sync

2. Run the server directly from the repository (requires uv to be installed):

uv run --directory /path/to/kill-process-mcp kill_process_mcp.py

Configure your MCP client to point at the server using the appropriate launcher configuration (see the README for examples for Claude Desktop and Cursor).

Tips and notes:

• The server relies on Python libraries mcp and psutil; ensure they are installed if using the manual setup (uv sync will install dependencies as needed).
• When using uvx, the client configuration should reference kill-process-mcp@latest to fetch the newest version automatically.
• If you encounter build or wheel issues after a Python update, a clean rebuild often fixes compatibility problems (e.g., delete virtual environments and run uv sync again).
• For safety, always verify the target process before issuing a kill command. The tools are powerful and can terminate critical system processes if misused.
• You can customize MCP client config to point to a local shim (kill-process-mcp) for faster startup and offline usage.