Works With Your Agentic Stack
PyRapide integrates with the frameworks you already use. Each adapter hooks into the framework's native observability system — callbacks, hooks, event handlers, or subclass wrappers — so your existing code stays unchanged.
AutoGen (Microsoft)
InterventionHandler on the runtime
- • Agent messages
- • Tool calls
- • LLM responses
- • Handoffs
"Trace causality across conversational multi-agent systems"
LangGraph (LangChain)
BaseCallbackHandler with run_id correlation
- • LLM requests/responses
- • Tool calls
- • Chain steps
"Add the 'why' to your stateful agent graphs"
CrewAI
Global tool hooks + step/task callbacks
- • Tool calls
- • Agent steps
- • Task completions
- • Delegation
"See the causal chain behind crew collaboration"
LlamaIndex
Root Dispatcher event handler (25+ event types)
- • LLM calls
- • Tool invocations
- • RAG retrieval
- • Queries
"Trace from query to answer through every retrieval step"
Agno (formerly Phidata)
Pre/post hooks + RunEvent stream
- • Full run lifecycle
- • Tool calls
- • LLM interactions
- • Team coordination
"Causal traceability for RAG-capable agents with memory"
OpenAI Swarm
InstrumentedSwarm subclass (drop-in replacement)
- • Run lifecycle
- • Tool calls
- • Agent handoffs
"See why your agents hand off and what happens next"
MetaGPT
Monkey-patch on Role.run, Action.run, Environment.publish_message
- • Role steps
- • Action executions
- • Inter-agent messages
"Trace the causal chain through your AI software company"
FlowiseAI
REST API + SSE event stream (no Python code needed)
- • Run lifecycle
- • Tool usage
- • LLM streaming
- • Agent reasoning
"Add causal tracing to visual agent workflows via API"
Ready to Add Causal Tracing?
Get started in under five minutes. Install PyRapide, pick your adapter, and see causal event graphs from your very first run.
Get Started