See which AI agent scenarios your real traffic actually covers.

Mibo connects the scenarios you define to passive production traces, separates scenarios with traffic from scenarios sitting idle, and surfaces unmatched conversations as evidence for where your test suite may need to grow.

Get started free Read the passive testing docs

A test suite can look healthy while real conversations go untested.

Synthetic scenarios tell you what you chose to exercise. Production traffic reveals what people actually ask. Mibo keeps those signals distinct so an observed match is useful evidence—not a claim that every possible interaction is covered.

Covered scenarios

A conditional scenario is covered when classified production traces match it in the selected window. Its traffic count is not a pass rate or a completeness score.

Idle scenarios

Defined conditional scenarios with no matching traffic stay visible as no traffic. They may be important edge cases—or simply not observed during the window.

Unmatched traces

A classified trace that matches no conditional scenario is a concrete gap signal. Review it and decide whether the conversation deserves a new test.

Turn real traffic into a sharper test backlog

Mibo connects scenario definitions and traces without adding evaluation to the customer request path. Use the coverage signal to decide what to investigate next.

Define the behavior

Create active test cases with specific scenarios and assertions. Use Run when relevant for situation-specific production checks; keep invariants and deliberate cases in their appropriate passive modes.

Send passive traces

After a real interaction, your system sends a canonical trace to Mibo through OpenTelemetry or Your API. Mibo stores it and schedules evaluation asynchronously against applicable active tests.

Find the next gap

Review Scenario Coverage to separate conditional scenarios with traffic, scenarios with no traffic, and classified traces that matched nothing. Use the evidence to prioritize a new or refined test.

Mibo makes the coverage signal honest and actionable

Covered is not passed

Coverage means a classified trace matched a conditional scenario. Open the passive result to review the semantic and procedural assertions; traffic presence alone does not say the agent behaved correctly.

Unmatched is not failed

No match means the classifier found no conditional scenario for that trace. Mibo keeps it separate from assertion failures and identifies cases to consider when expanding the suite.

Frequently asked questions

What does covered mean in Mibo?

A conditional scenario is covered when one or more classified production traces match the scenario during the selected time window. It describes observed traffic, not whether the assertions passed and not complete coverage of every possible interaction.

How does Mibo connect agent scenarios to real traffic?

Your system sends a trace after a real interaction through OpenTelemetry or Mibo's Your API path. Mibo compares the trace input with the scenarios in active conditional test cases, evaluates applicable tests asynchronously, and reports the matching scenario in Scenario Coverage.

What does an unmatched trace tell my team?

It tells you that a classified trace matched no conditional scenario in the selected window. That is a prompt to review the conversation and decide whether to add or refine a test—not proof that the agent failed.

Does Scenario Coverage prove my agent is fully tested?

No. It reports observed classified traffic for conditional scenarios in a time window. Empty inputs, unavailable classification, active-only checks, unobserved edge cases, and traffic outside the window can all limit what the signal tells you. Pair it with active runs and the evidence in each passive result.