Agenta

Open-source LLMOps platform covering the agent-reliability lifecycle: observability, prompt management, evaluation, auto-optimization. Founded by mahmoud-mabrouk.

Positioning

End-to-end tooling for teams building production LLM apps who need to: - Instrument traces (observability) - Manage prompts as versioned artifacts - Run offline/online evals - Auto-optimize prompts + judges via algorithms like gepa

Used in the "Judge the Judge" demo to instrument GEPA experiments and inspect generated candidate prompts mid-run.

Cross-references

• mahmoud-mabrouk — founder
• llm-judge-calibration — the workflow Agenta is tooling for