CrewAI

MeshFlow wraps CrewAI crews with a callback handler that emits per-agent and per-task spans. Governance, context packs, and the traceability ledger all work exactly as they do for native MeshFlow agents.

Install

bash

pip install "meshflow[crewai]"

Callback handler

python

from crewai import Crew, Agent, Task

from meshflow.adapters.crewai import MeshFlowCrewCallback

from meshflow.cloud import MeshFlowCloud

cloud = MeshFlowCloud(api_key="mf-...")

crew = Crew(

agents=[researcher, writer],

tasks=[research_task, write_task],

callbacks=[MeshFlowCrewCallback(cloud=cloud)],

)

result = crew.kickoff(inputs={"topic": "AI governance trends"})

Full example

The callback emits llm_call spans for model calls, tool_call spans for tool use, and a summary step span per task. Replay, approval gates, and governance policies apply automatically once spans are ingested.

Cost governance

Combine with CostCapGuardrail on the MeshFlow cloud side to enforce per-crew spending limits independent of the crew definition.

LangGraph

MCP Tools

Docs/Integrations/CrewAI