In our own software factory, AI agents build software – under our supervision. Agents based on Claude Code plan, implement, test and ship fully specified stories on their own.
Several specialized agents work together: for refinement, implementation, QA and acceptance, and design. Supervisor agents decide escalations in the interest of the product. Every decision is logged traceably and analyzed on a dashboard.
We assure quality independently of the coding agent: black-box and acceptance tests are written separately from the code and check the result from the outside.
The factory is built in Python and integrated with GitLab, Jira and Slack. Deployments run automatically via AWS CDK and TestFlight.
Achievements:
- Stories flow from refinement to delivery through AI agents
- Independent black-box and acceptance tests outside the coding agent
- Traceable agent decisions with logs and a dashboard
- Automated deployment of web and app
- Claude Code
- Multi-agent orchestration
- Prompt & Context Engineering
- Python
- AWS CDK
- GitLab
- Jira
- Slack
- TestFlight