Integrations / AutoGen
Trace AutoGen multi-agent chats
To trace AutoGen with Uncoil, wrap the conversation in trace_agent and record each tool execution as a step, or send AutoGen's OpenTelemetry spans to Uncoil over OTLP.
Your code
AutoGen
Uncoil
One run.
- 01One trace per conversation
- 02Tool execution per agent turn
- 03Model calls with tokens and cost
- 04Loops and failed turns
Setup
A few lines.
python
import uncoil
@uncoil.trace_agent(name="autogen-team")
async def run_team(task: str):
with uncoil.span("team-run", type="agent", input=task):
return await team.run(task=task)Full details in the documentation. All integrations are listed on the integrations page.
- Can I use OpenTelemetry instead of the SDK?
- Yes. Uncoil accepts OTLP over HTTP, so any OpenTelemetry instrumentation AutoGen emits can be sent to your project.
