Free · Open source · Native macOS

Start, manage and understand every local project.

One exact lifecycle for your local apps, servers, compound stacks, and coding agents. Fresh ports in. Guesswork out.

Install for macOS View on GitHub

brew install --cask JakeMawson/tap/launchstation

Available now · Developer ID signed · Apple notarized

Full stack preview

Running

Launch Systems Lab

compoundverifiedweb

Launch rail/tmp/launch-systems-lab-marketing-20260727

Frontendpython3 -m http.server '${PORT}'

API/bin/sh -c 'while :; do sleep 1; done'

Database/bin/sh -c 'while :; do sleep 1; done'

Launch the product frontend, API worker, and local data service as one exact session.

Launch configuration

Working directory: /tmp/launch-systems-lab-marketing-20260727
Run details: The primary action receives a fresh managed port; supporting services remain private process groups.
Arguments for next launch or relaunch: For example: --device iPhone --mode debug

Active sessions

1 running

Primary sessionPrimary · simulated

Session ID: sess-primary
Primary endpoint: Running

Interactive fictional session ready.

01Local by designAuthenticated loopback API

02Exact ownershipManager ID + process identity

03Native macOSSwiftUI · macOS 13+

04Agent readyApp · CLI · installed skill

An actual launch sequence

The stack starts in order. It stops in reverse.

Define one service or the entire local system. Required actions become a single exact session, with resolved endpoints available to every action that follows.

ORDER 00 · PROCESS
Database
Start the local data service first. Keep it private to the compound session.
./start-development-database
ORDER 10 · READY
API
Allocate a fresh port, wait for health, then expose the exact endpoint downstream.
GET /health
ORDER 20 · PRIMARY
Frontend
Receive the API URL, start on another managed port, and open the user-facing target.
VITE_API_URL=${LAUNCH_STATION_ACTION_API_URL}

launcher.definition

launch --create "Shop full stack" "Start the local stack" \
  --directory "$PWD" \
  --type process --port auto \
  -- npm run dev -- --host '${HOST}' --port '${PORT}'

Fresh managed port Readiness bound

Close with evidence

A port number is not ownership.

Launcher records the exact manager, process group, PID birth identity, endpoint, and action definition used by every session. Close targets that record—never a familiar name.

GuessingKill whatever uses port 3000

Exact lifecycleClose manager ID 63cab677

If ownership cannot be proven, Launch Station stops and reports an orphaned session.

Launch Station showing a fictional running full-stack session and its exact managed launch history.

The local control plane

Built around the things local development usually forgets.

Commands are easy. Knowing what started, when it became ready, where it is listening, and what may safely stop is the product.

PROCESS

Exact process groups

Direct commands run with recorded process-group and PID birth identity. Launcher never guesses from a name.

PORT

Fresh managed ports

Allocate a clean loopback port, inject it into the command, and refuse to take over unrelated listeners.

READY

Honest readiness

A service becomes running only after its HTTP health check returns a successful response.

Deliberate extra instances

Keep one ordinary primary session and create additional exact sessions only when you explicitly ask.

AGENT

One launch language

The native app, CLI, and coding-agent skill all read the same authenticated local catalog.

HISTORY

Durable session records

Inspect exact commands, endpoints, logs, roles, service counts, and terminal states after a run ends.

Launch Station showing a fictional full-stack session beside its launcher configuration editor. — EDITABLE DEFINITIONRunning sessions keep the exact definition they started with. Changes apply to later launches.

Codex · Desktop + CLI Claude · Desktop + CLI

One installation per product

Give every agent the same verified launch language.

Launch Station installs or refreshes its bundled skill for Codex and Claude Code. Each product shares one managed skill path across Desktop and CLI, so the app and terminal follow the same launch, readiness, history, relaunch, and exact-close workflow.

LIVE DISCOVERYCodex normally detects skill changes automatically. Claude detects updates live unless installation just created its top-level skills directory.

Launch Station Agent integration settings showing the shared Codex and Claude Code skill installers for Desktop and CLI, plus a standalone SKILL.md download. — The app verifies the exact files and version it manages. Reinstalling preserves sibling skills; receipt-proven uninstall removes only Launch Station-owned files.

CONTEXT

Spend tokens on the work

Agents read the verified skill and launch_details.md instead of repeatedly rediscovering commands, ports, and shutdown rules.

HANDOFF

Run what they make

A finished site or app becomes one registered launch command with a server-derived destination you can open immediately.

CATALOG

Keep projects organised

Codex, Claude, the native app, and the CLI share project-qualified launchers, tags, history, and exact runtime notes.

CLEANUP

Leave ports in a known state

Fresh ports and exact manager IDs give every agent the same safe close path—without killing whatever happens to use a number.

Free and open source

Install with one command.

Homebrew installs the Apple-notarized app, CLI, and local service. Your launcher catalog and session history stay in Application Support across upgrades and uninstall.

TerminalmacOS 13+

brew install --cask JakeMawson/tap/launchstation

STATUS Published · Launch Station 1.3.12

FREE

No account. No subscription.

Launch Station is free software for local macOS development workflows.

OPEN SOURCE

Read the source.

The complete application source and MIT licence are public on GitHub. Website and local configuration files are kept separate.

NOTARIZED

Download Launch Station 1.3.12.

The direct release is Developer ID signed, Apple notarized, stapled, and the same payload used by the Homebrew cask.

Direct from your workflow

Help steer what Launch Station becomes next.

Leave a quick rating, a suggestion, or both. Feedback goes directly to the Launch Station team.

Before you launch

The practical questions.

What can Launch Station run?

Direct processes, zsh commands, native macOS applications, URLs or files, Expo, Simulator-aware tools, and ordered compound workflows with several services.

Does it stop processes by name or port?

No. It uses the exact manager ID or the recorded process group and PID birth identity. When ownership cannot be proven, the session becomes orphaned instead of risking the wrong process.

Is the API exposed to the network?

The local API listens on loopback and every request requires a per-daemon bearer token stored in a private metadata file. Launcher is not intended to be exposed as a remote service.

Is the Homebrew cask available now?

Yes. The published cask installs the notarized app, CLI, and per-user service while preserving your launcher catalog and session history across upgrades and uninstall.

Can coding agents use it too?

Yes. Launcher includes an installable workflow skill for Codex and Claude Code, plus a deterministic read-only launch_details.md inside each initialized project.

How does the agent skill installation work?

Launcher detects supported signed Codex and Claude products, installs or refreshes one verified skill path per product, and shares that installation across the product's Desktop and CLI surfaces. Codex normally detects changes automatically; Claude detects them live unless its top-level skills directory was just created.

What version of macOS is required?

Launch Station supports macOS 13 and newer.