Agent Debugging & Replay

Session-replay style observability for AI agents with time-travel debugging, cost tracking and first-class integrations into agent frameworks like CrewAI and AutoGen.

Simulation-first testing and replay for voice/chat agents, borrowing evaluation methodology from autonomous-vehicle testing.

Agent trace analysis and security scanning (incl. MCP tool-poisoning research), with an Explorer UI for debugging agent runs.

Open-source observability for long-running AI agents that captures LLM calls, tool use and browser actions for step-level debugging and replay.

Autonomous LLM-Ops engineer that traces, debugs and optimizes LLM pipelines.