Active testing
Point Mibo at your agent and deliberately run the happy paths, critical edge cases, and regressions your team already knows about.
- Validate expected behavior on demand
- Check known paths before release
Test your agent before and after release.
Run planned scenarios before you ship. Use the same tests to check real customer interactions in production.
Try the live check below. No signup required.
Send a test request. See what Mibo catches.
A support agent answers your request. Mibo checks its response and shows you what passed and what failed.
One test suite, before and after release.
Define good behavior once. Mibo applies it to deliberate test runs before release and to real interactions after your agent is live.
Passive testing
Send completed traces from real traffic. Mibo selects the relevant tests and evaluates what happened without calling your agent again.
- Discover behavior you did not anticipate
- Stay outside the customer request path
Frequently asked questions
What is the difference between active and passive testing?
Active testing runs scenarios you choose against your agent before release. Passive testing evaluates traces created by real user interactions after they happen. Both modes use the same reliability tests, so your definition of good behavior stays consistent.
What exactly happens in this experience?
This is passive testing in action. Your synthetic request stands in for customer traffic, a real support agent responds, and Mibo evaluates the resulting trace against every relevant reliability test. Nothing is prerecorded.
What does the Agent Reliability Score mean?
For this run, it is the pass rate across the Mibo tests that matched your request. In production, the score becomes more representative as Mibo evaluates more scenarios and real traffic over time.
Can Mibo keep testing after I ship?
Yes. Mibo evaluates incoming OpenTelemetry or HTTP traces in the background, so you can catch regressions in real agent behavior without adding Mibo to the request path.
What happens to my trace data?
Trace payloads are encrypted at rest, and product access is scoped to the account that owns them. Unauthorized requests cannot use the API to confirm whether a trace exists.
Can I use Mibo with my own agent?
Yes. Connect any agent endpoint for pre-release testing, then send its execution traces to Mibo for continuous evaluation in production.
Test your own agent.
Create your first tests, run them against your agent, and see exactly where its behavior needs work.
Test my agent free