An agent that only reads and writes code is guessing whether its changes work. To maintain many projects I need one that can boot the app, run migrations, hit the page and read the actual error.

So I built Knecht. A Self-hostable tool. You connect your GitHub repos and it starts each project in an isolated Docker sandbox, generated with DDEV. Inside, an agent can do what a developer does locally. Reproduce a bug, run migrations, open the page in the browser etc.

Workflows can be started by triggers (Jira tickets, GitHub issues, PRs..) and can run across all projects at once, mixing AI agent steps with deterministic steps, similar to n8n.

Knecht is in beta and I am looking for people who maintain many projects. Works best with DDEV projects. https://github.com/knecht-works/knecht-cloud https://knecht.works/updates/beta-tester