CrewAI Integration
ANTS Platform traces CrewAI crews: crew runs, agent steps, LLM calls, tool calls and errors, with token usage, cost and latency per agent. Your crew logic does not change.
Choose your integration
Native OTel (CrewAI Enterprise)
No code. Add an OpenTelemetry collector in CrewAI Enterprise that points at ANTS Platform and set two environment variables.
SDK + EventListener
For self-hosted CrewAI or full Python SDK control. Register the EventListener from ants-platform[crewai] in crew.py.
What gets captured
| Native OTel | SDK + EventListener | |
|---|---|---|
| Crew run (root observation) | Yes | Yes |
| Agent steps | Yes | Yes |
| Agent in the AI Agent Center | One agent per CrewAI agent role | One agent per crew (agent_name) |
| Task execution | Yes, as task spans | Task prompt recorded on the agent step |
| LLM calls with tokens and cost | Yes | Yes |
| Tool calls with arguments and output | Yes | Yes |
| Errors | Yes | Yes |
| Trace tags | framework:crewai, crew:<crew name> | framework:crewai, agent:<name>, process_type, crewai_version, num_agents, num_tasks, plus your own tags |
Cost is calculated by ANTS Platform from the model and token usage.
text
MarketIntelligenceContentPipelineCrew (root)
research_market_signals (task)
Senior Market Research Analyst (agent)
LLM call (gpt-4o-mini)
Search the internet with serper (tool)
draft_content_strategy
Content Strategist
ℹ
Running CrewAI Enterprise? Native OTel is the quickest path: no code changes and no package to install. Start with the Native OTel guide.