I’ve been experimenting with Claude Code to go beyond “AI as a copilot” and instead simulate a small team of AI agents working on software development tasks.
The idea was simple:
Instead of asking Claude to help with isolated snippets, I structured it into a workflow where different “agents” handle:
- code generation
- review & validation
- architecture decisions
- cost/governance constraints
All orchestrated through a Java / Spring Boot backend.
What I found interesting is that the real challenge wasn’t generating code — it was coordination, governance, and control over the system behavior.
In practice, the hard problems became:
- preventing agents from diverging in logic
- maintaining consistency across outputs
- controlling cost and iteration loops
- introducing human decision points at the right time
I documented the full setup, architecture, and lessons learned here:
https://www.rheorix.com/en/2026/05/19/how-i-built-a-team-of-ai-agents-with-claude-code/
→ Full code and repository on GitHub: https://github.com/rheorix/agentic-company
Curious if anyone else is experimenting with similar multi-agent setups — especially in production or near-production environments.
What patterns are you using for orchestration and governance?