mcproc

mcproc is an MCP server that centralizes and manages long-running background processes for AI agents and developers. It provides a unified CLI and MCP exposure so agents can start, monitor, and control development servers, build watchers, daemons, and other background tasks while you retain full visibility and control from the terminal. The server offers tools to start, stop, restart, and inspect processes, as well as robust log handling with search and filtering, and integration with version-managers for different runtimes. Once registered as an MCP server, AI agents can invoke start_process, stop_process, restart_process, list_processes, get_process_logs, search_process_logs, and get_process_status to orchestrate development workflows.

With mcproc, you get: (1) Unified process management across AI agents and CLI users; (2) Intelligent log capture, persistence, and powerful search with regex and time-based filtering; (3) Project-aware grouping and real-time log following; (4) Support for toolchains and version managers (nvm, asdf, mise, rbenv, etc.); (5) Convenient commands for starting daemons, cleaning a project, and inspecting process state. Developers can also monitor processes locally via the CLI using commands like mcproc logs -f and mcproc ps to keep visibility while the AI agents operate in the background.

Prerequisites:

• A supported system (macOS, Linux, or Windows with WSL) with a suitable Rust toolchain installed for building from source.
• protobuf compiler if building from source for any generated bindings.

Install via Homebrew (macOS and Linux):

# Add the tap
brew tap neptaco/tap

# Install mcproc
brew install mcproc

Build from source:

git clone https://github.com/neptaco/mcproc.git
cd mcproc
cargo build --release

# Optional: install to PATH
cargo install --path mcproc

If you install via the binary, you can skip the build steps and just execute mcproc commands directly once installed.

Notes and tips:

• mcproc is designed to be run as an MCP server so AI agents can manipulate long-running processes through the MCP interface. Register it with your MCP client as shown in the README example.
• To ensure readiness synchronization, you can use --wait-for-log patterns when starting processes; mcproc will wait for the specified log line before reporting the process as ready.
• Environment variables for started processes are supported via the -e/--env flag (KEY=VAL). Use this to customize runtimes per project.
• The tool supports multiple projects and automatic cleanup of child processes when stopping a parent process, as well as a dedicated daemon for background management.
• If you run into permission issues, ensure mcproc has appropriate read/write permissions in the directories used for logs and state (often within XDG base directories).
• When integrating with MCP clients, you can place mcproc into your mcpServers configuration as shown in the README example to enable discoverability and control from AI agents.