One product, three editions — the same codebase; the edition just sets what's switched on: Full (registers the install with a free account — no card — and gets the community Marketplace, plus the optional paid features — Autopilot Cruise, AutoCognition, Everywhere), Stand-alone (V1) (everything local, zero network calls), and the hosted Web Demo (try it in the browser, nothing to install). Free, open, local-first — Linux, ARM Linux, macOS, and Windows (via WSL2); every other device connects through the Everywhere relay.
Try the Web Demo edition first — no install
All PHP backends, frontend, shell scripts, documentation, and tool definitions in one tarball. Partway through install, you pick which of the two installable editions this machine runs (the third, the Web Demo, is hosted — try it here, no download):
.tar.gz above works
natively — tar has shipped built into Windows since 2018 and is
standard on macOS. Prefer no terminal at all? Use the
manual-install .zip
below instead — plain double-click extraction.
tar -xzf cockpit_install.tar.gz, then run
install_windows.ps1 from inside the extracted folder — it
sets up WSL2 and does the rest. Full steps in the
Setup Guide's Windows section.
Written and reviewed, but not yet run on a real Windows machine —
tell us how it goes.tar -xzf on the
.tar.gz from Terminal. See the
macOS section.
CI-verified — an unattended end-to-end install run passes on a real
Apple Silicon runner.install_cockpit.sh as regular Linux — tar -xzf then run the script.
CI-verified — an unattended end-to-end install run passes on a real ARM64 runner.Docker packages AI Cockpit and Ollama into one self-contained container (they share it — see the Setup Guide for why), so you skip the native/WSL2 setup entirely. Runs the same way on Linux, Windows (via Docker Desktop on WSL2), and macOS (via Docker Desktop's built-in virtualization).
Not yet published to Docker Hub — download the small bundle below
(Dockerfile + docker-compose.yml + setup
instructions), build it locally. Once it's running, use it exactly like a
normal local install: open your regular browser on the host machine and go
to http://localhost:8080 — you never interact with the
container directly after that. Written and reviewed carefully, but not yet
actually built/run end-to-end — let us know how it goes.
Don't want to run the install script at all? This .zip has the app's files already extracted, plus a step-by-step Manual Install Guide for Linux, macOS, and Windows — no script required, includes the one-line step for choosing Full account vs Stand-alone by hand. The install scripts are still included inside, in case you change your mind partway through.
Download cockpit_manual_install.zipAndroid, iOS, or anything else — connect to your own Cockpit from another device through the Everywhere relay. Free account with the Everywhere tier, nothing installed on the phone beyond a browser tab.
Open the Everywhere relayCard 1/2/3 each accept several CLIs — these are just the default choices below, not the only options (see Features).
npm install -g @anthropic-ai/claude-code)npm install -g @openai/codex); or Gemini's free-tier CLI (@google/gemini-cli)
The installer adds AI Cockpit to your applications menu (under an
“AI” category) plus the cockpit-start / cockpit-stop
commands. Launching shows a branded splash, then opens Cockpit in a clean app window.
Install both versions and AI Cockpit V1 (Stand-alone) gets its own menu
entry and cockpit-v1-start command — only one version runs at a time.
Step-by-step setup including whisper.cpp / Vulkan compilation, LibreTranslate systemd service, and first-run verification.
Read the Setup GuideComplete user manual — all sections, opens in-app too.
Full installation, configuration, and verification steps.
Hand-copy steps for every OS, no script — also inside the manual-install .zip.
Dockerfile, compose file, and setup steps — separate download, see above.
Also inside the tarball itself: MANIFEST.txt (SHA-256 checksums for
every file) and VERSION.