RunVouchGet a free key

Docs › AutoGen scheduled agent monitoring with RunVouch

AutoGen scheduled agent monitoring with RunVouch

AutoGen scheduled agent monitoring for the chat that keeps agreeing to continue: two AutoGen agents that keep agreeing to continue, or a tool that keeps failing and being retried by the assistant, look exactly like work until the bill arrives.

On this page
  1. How it runs on AutoGen
  2. Store the key
  3. Wrap the job
  4. Register the cadence and caps
  5. What goes silent on AutoGen
  6. What AutoGen does not tell you
  7. What RunVouch detects
  8. Set up in two minutes
  9. Also running agent frameworks?

How it runs on AutoGen

Around the chat or team run with the Python client, reporting tool executions through run.tool().

Store the key

An environment variable wherever the script runs.

Wrap the job

import asyncio, runvouch
from autogen_agentchat.teams import RoundRobinGroupChat

async def main():
    with runvouch.vouch("nightly-triage", evidence=lambda: {"labels": labels_applied > 0}) as run:
        team = RoundRobinGroupChat([triager, verifier], max_turns=12)
        result = await team.run(task="Triage new issues and apply labels")
        for m in result.messages:
            if getattr(m, "type", "") == "ToolCallExecutionEvent":
                run.tool("tool", {"content": str(m.content)[:200]})

asyncio.run(main())

rv fails open: if RunVouch is unreachable the job still runs and you get one warning line.

Register the cadence and caps

rv agent nightly-report --cadence 24h --grace 30m --max-runtime 1h --evidence --cap-run-cost 2

Register the agent once, from anywhere with the key. Cadence is what turns a schedule that stopped into an alert; --evidence makes a run without evidence a failure; the caps pause the agent when it overspends.

What goes silent on AutoGen

What AutoGen does not tell you

AutoGen ends a team run on a termination condition or at max_turns, and returns the message history either way; reaching the turn limit is a normal stop reason, not an error. Usage is available per message in the result. Nothing in AutoGen knows when the run was supposed to happen or whether the result was used.

What RunVouch detects

AlertWhat it means here
MISSEDcadence plus grace passed and no run started
FAILEDnon-zero exit or a reported failure, with the stderr excerpt
NO_EVIDENCEthe run said ok but the file, URL or assertion you required is missing
STALLEDa run started and never ended within max runtime
RETRY_STORMthe same tool called with identical input many times in one run
BUDGET_RUN / BUDGET_DAYcost cap crossed; the agent is paused until you resume it
DRIFTduration or output size far off its 7-run baseline

Set up in two minutes

  1. Get a free key (3 agents, no card) and store it where this page says.
  2. wrap the job: copy the snippet above into the scheduled job.
  3. Register the cadence once: rv agent nightly-report --cadence 24h --grace 30m --evidence, or let the first run create the agent and set the cadence on the dashboard.
  4. Send one test alert: curl -X POST https://api.runvouch.com/v1/settings/test-alert -H "X-API-Key: $RUNVOUCH_KEY". The next missed, failed or empty run reaches the same channels.

Also running agent frameworks?

OpenAI Agents SDK. Pydantic AI and smolagents. Dify and Flowise. Ollama and local LLM batch jobs. The full list of runtimes is on integrations, and the same agent can report from several of them: one job, one cadence, wherever it happens to run.

Need a key? Get a free key · Stuck? contact