Beta

CrewAI Integration

Capture role-based CrewAI task orchestration as trace sessions to compare team-of-agents cost/latency tradeoffs in Neurovn.

Install & Run

Install

Setup

pip install crewaipip install neurovn

Run

Execute

NEUROVN_API_URL=https://agentic-flow.onrender.com python crew_run.py

Architecture Flow

Roles

Map each Crew role to a decorated agent function.

Tasks

Map external calls/handlers to decorated tool functions.

Review

Send sessions to Neurovn and compare scenarios across model mixes.

Implementation Snippets

Pattern

python

from neurovn import trace@trace.agent(name="Research Crew Agent", model="Claude-4-Sonnet", provider="Anthropic")def research_task(topic: str) -> str:    return "brief"

Troubleshooting

Keep crew role names stable so historical comparisons remain readable.

Wrap the crew kickoff entrypoint in `with trace.session(..., source="decorator", canvas_name="...")` so one crew execution maps to one named canvas.

Use branch-specific wrappers when crew tasks diverge by condition.

Validate provider/model combinations against backend pricing registry.

Related Integrations

code to-canvas

Integration guide

langchain

Integration guide

langgraph

Integration guide

Backend contracts: `/api/estimate`, `/api/traces/sessions`, `/api/canvases`