AI coding agents build apps in the terminal. Floo deploys them from the terminal. Here's how to wire it up.
Install
curl -fsSL https://getfloo.com/install.sh | sh
This installs the floo CLI. Works on macOS and Linux.
Set up skills
Floo ships with agent skills that teach your coding agent how to use the platform. From your project directory:
floo skills install
This drops a skill file into your project that your agent picks up automatically. It covers deploying, managing env vars, viewing logs, and everything else — so your agent doesn't need to guess at commands.
Log in
floo auth login
For headless environments (CI, cloud VMs, containers), set an API key instead:
export FLOO_API_KEY="floo_..."
That's the entire setup. Three steps.
Deploy
From your project directory:
floo deploy
Floo detects your framework, builds a container, and deploys it. You get a live URL back in seconds. No Dockerfile, no build config, no dashboard.
Every agent can parse the output with --json:
floo deploy --json
Two ways to deploy
Local deploy — run floo deploy from the terminal. Your agent builds the app and deploys it in the same session.
GitHub push — connect your repo through our GitHub App and every push to your main branch triggers a deploy automatically. Set it up once:
floo github connect
Use whichever fits your workflow. Both produce the same result.
Environment variables
floo env set DATABASE_URL="postgresql://..."
floo env set STRIPE_KEY="sk_..."
Encrypted at rest, injected at runtime. No .env files in git.
Beyond deploy
Floo also handles custom domains, deploy logs, rollbacks, and other parts of the software lifecycle — all from the CLI, all agent-friendly. We'll cover those in future posts.
If your agent can run shell commands, it can ship production apps with Floo.