NO SETUP REQUIRED
Say what you want — a timeclock, a booking form, a tracker for your guitar pedals. Zeltro gets everything ready behind the scenes, your AI builds it, and minutes later it’s running on your own computer.
THE DESKTOP APP
Everything the CLI does, without typing a command — create with AI, install from the app library, start and stop services, and see every project on the machine at a glance.
Agent terminals are built in. AI sessions run in real tabbed terminals inside the window, so you can watch your agent work and jump in whenever you want.
Download
Two pieces. The CLI does the work; the GUI is an optional desktop front end. Neither ships as a package — both install from source, so what you install is what runs, and updating is a git pull.
Straight to it? The steps are below — the GUI is optional.
Every platform, and what differs on each.
Zeltro does three things, and they reinforce each other.
One zeltro-postgres, one zeltro-redis, one
zeltro-mariadb, used by every project on the machine.
Run ten projects, you've still got one of each. Cross-project hostnames
resolve from the browser and from inside containers, so a Laravel
app and a FastAPI service can share a database with zero networking setup.
zeltro install grafana — fully configured,
running, reachable at http://grafana/ in under two minutes.
Or scaffold a fresh Laravel, FastAPI, Django, or Express project with
zeltro new. No env-var spelunking. No
"why isn't db:5432 resolving" debugging. No copy-pasting
docker-compose files from someone's gist.
Zeltro's command surface is shaped around AI-driven development. Your agent gets a stable platform — shared services, hostname routing, three known runtime images — so it focuses on the actual app instead of reinventing the runtime. Works with Claude, Codex, and Gemini.
WHY ZELTRO
You can. It'll just do it slowly, expensively, and a little differently every time. Zeltro is the durable substrate underneath — with or without an agent driving it.
http://django-app/ from your browser and from your Express container.zeltro update-installer regenerates installers when upstream drifts. The platform self-maintains.Zeltro isn't competing with your AI agent. It's the workshop your agent works in — tested tools, shared resources, predictable setup — so the agent (and you) skip rebuilding the workshop every time.
BUILT TO SAVE AI TOKENS
Framework scaffolding, container networking, shared services, secret generation, and 200+ OSS app installs are all pre-baked. Your agent doesn't burn tokens rediscovering how to wire nginx + php-fpm, what env vars Mastodon needs, or how to generate VAPID keys every session.
Faster results. Cheaper bills. Fewer "ugh, why isn't this working" loops.
$ zeltro install grafana
Installing Grafana...
Configuring shared Postgres database...
Starting container...
Grafana is ready! (HTTP 302)
URL: http://grafana/
$ zeltro create "A timeclock app in Django"
AI agent scaffolding project...
- migrations, models, views, templates
- shared Postgres connection wired
Running at http://timeclock/
No Docker knowledge required. No config files to wrestle with.
Install a curated open-source app, fully wired up to shared services, in under two minutes.
$ zeltro install grafana
$ zeltro install gitea
$ zeltro install n8n
$ zeltro install nextcloud
$ zeltro install mastodon
200+ apps. Pinned versions. Tested end-to-end on a multi-machine fleet.
Describe a project in plain English. Zeltro builds and wires the whole environment; your AI agent only customizes it.
$ zeltro create "A timeclock for employees in Django"
$ zeltro create "A customer check-in system in Laravel"
$ zeltro create "An inventory tracker in Express"
$ zeltro create -f detailed-spec.md # or pipe a big prompt in
Migrations, models, routes, templates — all wired to a real database, all running locally. Big prompt? Use -f file, < file, or pipe it in.
Four steps. From zero to a real app on a local URL.
This is Zeltro itself — the command line tool that does all the work. Pick your platform; one curl, no manual setup.
Choose your platform
curl -fsSL https://raw.githubusercontent.com/CaneBayComputers/zeltro-cli/master/install-ubuntu.sh | bashInstalls the CLI only. The desktop app is optional and comes later — see step 5.
Picks up Docker group membership and PATH changes from the installer.
sudo rebootSets up your projects directory, Docker network, hosts entries, and AI agent. Safe to re-run — existing values stay as defaults.
zeltro configurePick one of three paths — whichever fits the project you have in mind.
zeltro install <app> # click an app below for the exact commandzeltro create "A timeclock for employees in Django"zeltro create "https://github.com/your-org/your-repo"Clones the repo, adapts its docker-compose.yaml to use Zeltro's shared databases, and starts everything. Your AI agent reads the project's README and picks up where the upstream left off.
zeltro new <type> my-app
Everything above is the CLI and is all you need. Zeltro GUI is a separate, optional desktop front end for it — same projects, same services, just visible and clickable.
Install the GUI
One command, the same shape as the CLI above — it clones the repo for you and builds from source. On Linux and macOS it installs the CLI first if zeltro is missing.
curl -fsSL https://raw.githubusercontent.com/CaneBayComputers/zeltro-gui/master/install-ubuntu.sh | bashSwap for install-fedora.sh, install-arch.sh or install-mac.sh. It pulls the npm dependencies, compiles the TypeScript and rebuilds the native terminal module against Electron — a few minutes, mostly npm install. Re-running it is safe, and running it from inside an existing checkout installs that checkout instead of cloning.
Windows works differently. The GUI runs natively, but there is no local Zeltro for it to drive — it connects over SSH to machines that have one, added under Settings → SSH Hosts. Right-click PowerShell, choose Run as administrator, then: irm https://raw.githubusercontent.com/CaneBayComputers/zeltro-gui/master/scripts/install-windows.ps1 | iex Full details →
No licence, no tiers, no paywalled features — both the CLI and the desktop app are MIT and open source. If it saves you time, you can chip in. Entirely optional. Read why it exists →
Click an app to see the install command. All wired to shared databases. All reachable at http://<app-name>/.