LangChain / LangGraph
Callback handler integration for LLM/tool lifecycle visibility in existing chains and graphs.
pip install agentdbg[langchain]
Optional adapters over a framework-agnostic core
Integrations for common Python agent stacks.
Use AgentDbg directly in custom code, or enable thin adapters so framework callbacks map to the same event schema and timeline.
OpenAI Agents SDK
TracingProcessor integration for generation spans, function calls, and handoffs.
pip install agentdbg[openai]CrewAI
Execution-hook adapter with active run context so timeline events stay tied to current debugging sessions.
pip install agentdbg[crewai]Integration principles
Every integration is designed to keep the core small, stable, and portable.
What integrations should do
- Remain optional dependencies.
- Map framework events to existing AgentDbg event types.
- Record events only within active AgentDbg run context.
- Degrade clearly when dependencies are missing.
What they should not do
- Create new event taxonomy just for one framework.
- Force framework lock-in.
- Add hidden cloud requirements.
- Change the local-first debugging workflow.
Start with core tracing, then add adapters where they save time.
The same timeline model works across custom code and framework integrations.