Pitchfork manages the long-running commands your project needs: an API, a database, a frontend server, or a background worker. Define them in pitchfork.toml, then start, inspect, and stop them from any terminal. A background supervisor keeps track of the processes and their logs after the CLI exits.

Get started

Install with mise:

mise use -g pitchfork

You can also use cargo install pitchfork-cli --locked or download a binary from GitHub releases. See installation for platform details and shell completion.

Try a daemon without a config file (requires Python 3):

pitchfork run demo --port 8000 -- \
  python3 -u -m http.server 8000 --bind 127.0.0.1
pitchfork status demo
pitchfork logs demo --tail
# Ctrl+C leaves the daemon running. Stop it when you're done:
pitchfork stop demo

Open localhost:8000 while the daemon is running.

Put your services in version control

Create pitchfork.toml in your project. This example assumes Redis is installed and your project has a server.js that reads PORT and REDIS_URL:

#:schema https://pitchfork.jdx.dev/schema.json

[daemons.redis]
run = "redis-server --port $PORT"
port = 6379
ready_cmd = "redis-cli -p $PORT ping"

[daemons.api]
run = "node server.js"
port = 3000
depends = ["redis"]
env = { REDIS_URL = "redis://127.0.0.1:{{ daemons.redis.port }}" }
ready_http = { url = "http://127.0.0.1:3000/health", timeout = "30s" }
retry = 3

pitchfork start api       # Starts Redis first, then waits for the API's health check
pitchfork start --local   # Starts all daemons in the project's merged local config
pitchfork list --project  # Shows this project's daemons
pitchfork restart api     # Applies config changes and restarts the API
pitchfork stop --local    # Stops local services in reverse dependency order

Starting an already running daemon leaves it running. Independent dependencies start in parallel. --all includes global daemons as well as local ones.

Make it fit your workflow

When you need to…	Use…
Start services when you enter a project	Shell hooks with auto = ["start", "stop"]
Wait for a service to accept requests	Ready checks using output, HTTP, TCP, or a command
Recover from a crash or failed health probe	Retries and health checks
Restart after a source edit	File watching with glob patterns
Keep a stable URL across port changes	Port assignment and reverse proxy
Inspect services and their output	pitchfork tui, the web UI, and structured logs
Run tasks on a schedule	Cron scheduling
Connect an AI assistant	The built-in MCP server

The docs also cover namespaces and worktrees, mise integration, lifecycle hooks, and resource limits.

Contribute

See CONTRIBUTING.md for setup, checks, and documentation development. Found a problem? Include your version, config, and relevant logs in a GitHub issue.

Sponsors

Sponsored by

   

View all sponsors

Contributors