Ship your first app¶
From nothing to a live HTTPS URL. Realistically five minutes, most of it waiting for a Docker build.
You'll need: a GitHub account (that's how you sign in), and Docker on your machine — the deploy builds your image locally rather than on the box.
1. Sign in¶
Go to allocus.dev and sign in with GitHub. Sign-in asks only
for the read:user scope, and only your username is kept.
You'll then be asked for an email address, once, before the dashboard opens. This isn't a marketing capture — GitHub sign-in doesn't hand over an address, and Allocus needs one so it can tell you about a security incident, a price change, or the service shutting down. Those are commitments in the Terms, and one of them is a legal obligation, so the dashboard genuinely won't proceed without it. It's used for nothing else.
2. Launch your box¶
From the dashboard, start your subscription and hit Launch my box. Provisioning
takes a minute or so: a VM in Paris, Docker, the edge proxy, and a wildcard TLS
certificate for *.<you>.allocus.dev, so your first app is HTTPS-valid the moment it
lands.
While it warms up, copy your API key. It's shown once — only a hash of it is stored, so it can't be shown to you again. If you lose it, revoke it from the dashboard and make a new one.
3. Install the CLI¶
One line installs the allocus CLI:
Then log in with the key from your dashboard:
4. Add the Claude plugin¶
The CLI can deploy on its own — skip to deploy by hand if that's all you want. The plugin is what makes "deploy this to allocus" work, and it installs separately from the CLI:
5. Deploy¶
Open your app's repo in Claude and say:
The skill writes a Dockerfile and an allocus.yaml if you don't already have them,
then runs the deploy. A moment later your app is live at:
That's the whole loop. Deploy another app the same way — it lands on the same box, at its own subdomain, with its own working HTTPS, and costs nothing extra.
Rather not use Claude? The CLI does the identical thing:
allocus init then allocus deploy. See deploy by hand.
If it didn't work¶
The most common first-deploy failure by a wide margin: your app is listening on
127.0.0.1 instead of 0.0.0.0. Inside a container, localhost means "nobody else can
reach me", so the proxy gets a connection refused.
Two commands tell you nearly everything:
allocus deploy --wait # blocks and names what failed, instead of just "queued"
allocus logs # the container's own output
When something's wrong covers the rest by symptom.
Worth reading next¶
- How it works — what just happened, end to end. Same pipeline whether Claude drove it or you did.
- Apps with a database — a frontend, a backend and Postgres deployed and rolled back as one unit.
- Your data — where it lives, who can read it, and the part where your box is not backed up for you. Read that one before you put something you care about on it.