open source agent engineering platform
PandaProbe is an open-source agent engineering platform that gives you deep observability into AI agent applications. Use it to trace, evaluate, monitor and debug your AI agents in development and production.
PandaProbe provides SOTA metrics purpose-built for agent behavior, LLM-as-judge scoring with structured feedback, and session-level evaluation (not just isolated traces). This allows teams to score and pinpoint exactly where agents drift across entire lifecycles.
PandaProbe captures full agent trajectories (tool calls, LLM hops, decision branches) and applies state-of-the-art metrics to detect agent uncertainty over long trajectories. This enables developers to spot issues and behavioral drift before users experience problems.
PandaProbe offers one-line instrumentation for major agent frameworks (LangGraph, LangChain, CrewAI, Google ADK, Claude SDK, OpenAI SDK) and works with any LLM provider out of the box. Spans and metadata are captured automatically, reducing integration overhead.
PandaProbe is open source (Apache 2.0) and self-hostable with all core features available free. Teams can deploy on their own infrastructure for unlimited scale, or use the managed cloud option and upgrade only when team size or trace volume requires it.
PandaProbe schedules eval runs against production traffic on custom cadences (daily, hourly, or cron), alerts on metric regressions across agent versions, and catches behavioral drift the moment it appears. This enables proactive regression detection without manual intervention.