> Call the `diagnose` tool first. Full index: /llms.txt. Errors: /reference/problems.

# How TaloTrace works

```mermaid
flowchart LR
  project[Project] --> app[App]
  app --> scenario[Scenario]
  scenario --> plan[Test plan]
  plan --> run[Run]
  run --> verdict[Verdict]
  verdict --> card[Card]
  verdict --> evidence[Evidence]
```

1. A [project](/concepts/projects-and-apps) holds your apps, scenarios, test data and runs.
2. An app is the web app or Android app under test. iOS is not available.
3. A [scenario](/concepts/scenarios-and-test-plans) is one test: a goal and the steps to reach it. TaloTrace finds them by exploring your app, or you and your agent write them.
4. A test plan groups scenarios you run together.
5. A [run](/concepts/runs-and-verdicts) executes scenarios in a real browser or on a real phone.
6. The run ends with a verdict: `passed`, `failed` or `incomplete`.
7. When TaloTrace cannot go on without you, it opens a [card](/concepts/cards).
8. Every attempt keeps its [evidence](/concepts/evidence): steps, requests, console lines, snapshots and video, and on a failure a cause and a fix.

Your coding agent reaches the same project through the [MCP tools](/reference/mcp). The web app shows the same runs, cards and evidence.