Free + Open Source · Apache 2.0

open source
agent engineering platform

Traces, evals, and metrics to debug and improve your AI agents.

Get Started View on GitHub

Self-hostable · No vendor lock-in · Built for scale

Core Capabilities

Everything your agents
need to reach production.

A unified platform for the full agent development lifecycle — from first run to continuous improvement.

Observability

Capture every trajectory.

Deep-dive into agent reasoning paths with high-fidelity traces. See every tool call, sub-agent call, LLM invocation, and decision branch in real time.

Evaluations

Automated risk scoring.

Run critical episode analysis to detect where agents fail before they reach production. Score traces against custom rubrics or golden datasets.

Metrics

Actionable insights.

Monitor trajectory metrics and alert your agents before issues arise. Spot regressions and bottlenecks before they become incidents.

For Developers

Built for the agentic era.

Integrate in seconds with our Python SDK.

Works with any agent framework
Free hobby plan to get you started

pip install pandaprobe

Read the Docs Quickstart →

agent.py

python

1	import pandaprobe
2
3	# Initialize the PandaProbe client
4	probe = pandaprobe.Client(
5	api_key="pp_live_xk9mZ2...",
6	project="my-agent",
7	)
8
9	# Wrap any LLM call with automatic tracing
10	@probe.trace(name="research_agent")
11	def run_agent(query: str):
12	response = llm.invoke(query)
13	probe.log_eval(response, rubric="factuality")
14	return response