spm
CLI and MCP server to run native systemd, launchd and Windows services for coding agents
- Pricing
- Open source
- Platforms
- Linux, macOS, Windows
- Version
- 0.4.2
- Listing updated
- 1 day ago
Similar MCP servers
About spm
spm is an open-source MCP server for developer tools. It is available on Linux, macOS, Windows, CLI and self-hosted.
The README says spm gives one command-line interface for running and inspecting long-lived processes as real native services on Linux (systemd), macOS (launchd) and Windows (Windows Services). Instead of leaving a dev server or worker in a background shell with & or nohup, the process is registered with the operating system's own service manager, so it can restart on crash and at login, and its logs remain readable.
The project is aimed at coding agents such as Claude Code, Codex and Cursor, which can reach spm through a bundled stdio MCP server (spm mcp) and an agent skill. Human operators can use the same commands from a shell, and the PM2 command mapping in the README suggests it is meant as a replacement for PM2 in fork mode.
Services default to the current user, with a --system scope for machine-wide services that needs root or Administrator rights. Scheduled jobs, health checks, readiness probes and resource limits are covered in the README, with platform differences noted, such as some health and resource options being Linux-only on Windows.
What it does
- Native service registration
- Creates services through systemd, launchd or Windows SCM and Startup integration so processes survive logouts and reboots.
- Agent MCP server
- Ships a local stdio MCP server with structured tools for listing, inspecting, reading logs and controlling services.
- Bounded log reading
- Reads service logs with line limits, time filters and regex grep, and separately shows spm's own run and exit records.
- Scheduled jobs
- Runs one-shot jobs on cron or interval timers with catch-up and randomized delay options.
- Declarative spm.toml
- Validates, diffs and applies a TOML file describing several services, with platform-specific override blocks.
- Health and readiness checks
- Waits for HTTP, TCP or command readiness at start and can restart or stop a service after failed periodic health checks.
Good for
- Keep a dev server or worker running after a coding agent's session ends
- Let an AI coding assistant start and inspect background services through MCP
- Replace PM2-managed Node.js processes with native OS-managed services
- Run recurring backups or sync scripts on a schedule with native timers
- Manage a group of related services from one reviewed configuration file
Pricing
Open source
MIT licensed open-source CLI; no paid plan mentioned.
spm FAQ
Which operating systems does spm support?
The README describes Linux with systemd, macOS with launchd and Windows with Windows Services. Some features, such as health timers, instances and resource limits, are documented as Linux-only on Windows.
Does spm require root to run services?
No. User services are the default and do not need elevation. The --system flag selects machine-wide services, which require root on Linux and macOS or an Administrator terminal on Windows.
How do coding agents connect to spm?
The README lists commands for Claude Code (plugin or mcp add), Codex (codex mcp add), Cursor and VS Code (install buttons or mcp.json) and Claude Desktop, all running spm mcp over stdio.
Is spm a PM2 replacement?
The README includes a PM2 command map and says PM2 cluster-mode socket sharing, watch mode and remote management are not implemented.
Is spm free?
Yes. spm is open source, so you can use and self-host it for free.
What are the best alternatives to spm?
The closest alternatives to spm are Linux MCP daemon (mcpd), VPS MCP & Digital Secretary, devglow, pulsar and Plug and Ship.
Where this comes from
- Description: written from the project's README and its MCP Registry entry.