I Give My AI Agent a Git Link and Get Back a Live App. Here's What It Does When the Build Breaks.

Description and use

Automating the deployment loop: an AI agent creates a service, triggers a build from a git repo, streams logs, parses failures, commits fixes, rebuilds, and verifies the live URL.

Problem

Automating the deployment loop for a Git-hosted app so an AI agent can build, detect failures, fix code, and redeploy without human intervention.

Outcome

The agent can fully automate the deployment loop, handling failures by reading logs, applying fixes, and redeploying until the app is live.

Workflow

• Create a service with placeholder image "pending" and port 3000 via POST /v1/services
• Start a build from a git URL with source_type "git", source_url, and source_ref (commit SHA) via POST /v1/services/{svc}/builds
• Stream the build output with GET /v1/services/{svc}/builds/{build}/logs?follow=true
• When build fails, retrieve the full log with GET /v1/services/{svc}/builds/{build}/logs and status with GET /v1/services/{svc}/builds/{build}
• Parse the last 40 lines of the log to identify the error
• Fix the code (e.g., run npm install, commit lockfile), push to repo, and start a new build with the new SHA
• After deployment success, check service status and URL with GET /v1/services/{svc} and request the URL to confirm it answers

Evidence

agentic-hosting is a single program called ah that runs on your server. Agents talk to it over an HTTP API.

Source

Create a service, the server's name for one app.

Source

Start a build from a git URL.

Source

Stream the build output.

Source

If it fails, read the end of the log.

Source

Fix the code, push, and build again.

Source

When it's running, check the URL.

Source