> ## Documentation Index
> Fetch the complete documentation index at: https://checksum.ai/docs/llms.txt
> Use this file to discover all available pages before exploring further.

# Deep vs Standard Modes

> Every Checksum pipeline (detection, generation, and healing) runs in one of two modes. Standard is fast and direct. Deep interviews you, builds a knowledge base of your app, and plans before it implements. The mode sets how thorough the agent is and how much input it needs from you.

## At a glance

| | Standard <span className="pill">default</span> | Deep |
| - | - | - |
| **Speed** | Faster turnaround | Slower, more thorough |
| **Your involvement** | None. Fully autonomous from start to finish. | You answer interview questions and approve the plan, then it runs on its own |
| **Parallelism** | Start 10 sessions and walk away | Approve each plan, then leave them to finish |
| **Best for** | Well-defined flows, straightforward apps, fast iteration | Complex apps, first-time coverage, business-logic-heavy flows |

## Standard mode

Standard mode is **faster and more direct**. The agent goes straight to implementation without an extended planning phase.

```text theme={null}
Implementation → Review → Checksumify → Verify
```

**Best for:**

* Well-defined test flows with clear steps
* Straightforward applications
* When you want faster turnaround
* Iterating quickly on test coverage

## Deep mode

Deep mode adds **planning, knowledge-base analysis, and a more thorough review**. The agent interviews you first, builds an understanding of your app, and then implements with that context.

```text theme={null}
Interview/Plan → Knowledge Base Update → Implementation → Review → Checksumify → Verify
```

**Best for:**

* Complex applications with nuanced workflows
* First-time test generation on a new app
* Flows that need a deep understanding of business logic
* When quality and coverage matter more than speed

## Pipeline stages

| Stage | Standard | Deep | What happens |
| - | - | - | - |
| **Interview/Plan** | — | ✓ | The agent asks questions about your app and writes a detailed plan. The session enters *Waiting for Input*, then *Waiting for Approval*. |
| **Knowledge Base Update** | — | ✓ | The agent builds an understanding of your app's architecture and patterns |
| **Implementation** | ✓ | ✓ | The agent writes the test story and the Playwright code |
| **Review** | ✓ | ✓ | The agent reviews its own work for quality, correctness, and best practices |
| **Checksumify** | ✓ | ✓ | The agent validates the tests against the Checksum CLI |
| **Verify** | ✓ | ✓ | The agent runs the tests to confirm they pass |

<Frame caption="Standard skips the interview and knowledge-base stages.">
  <img src="https://mintcdn.com/checksum/jreTwWrmFV2djRX_/images/pipeline-stages-diagram.png?fit=max&auto=format&n=jreTwWrmFV2djRX_&q=85&s=f79847f4bb9cfe7500c8c86646821229" alt="Diagram of Standard and Deep pipeline stages" width="1600" height="620" data-path="images/pipeline-stages-diagram.png" />
</Frame>

<Note>
  **Healing uses a different Standard pipeline**

  Standard **healing** runs `Triage → Fix` instead of the stages above. Deep healing uses the full Deep pipeline. See [Auto-Healing](/docs/auto-healing).
</Note>

## All pipeline variants

| Operation | Standard | Deep |
| - | - | - |
| **Test generation** | Fast generation, direct implementation | Thorough planning + generation |
| **Test detection with source code** | Quick code analysis | Deep code exploration with planning |
| **Test detection without source code** | Direct app exploration | Extended exploration with planning |
| **Auto-healing** | `Triage → Fix`: classify each failure as bug or test issue, then fix test-local issues | Root-cause analysis via the full Deep pipeline: plan the fixes, learn what changed in your app, then implement and verify |

## When to switch

| Scenario | Recommended mode |
| - | - |
| Generating tests for a simple CRUD app | Standard |
| First-time setup on a complex SaaS platform | Deep |
| Healing a few broken selectors | Standard |
| Healing tests after a major app refactor | Deep |
| Quick detection of obvious user flows | Standard |
| Comprehensive coverage of a feature area | Deep |

## Choosing a mode

You choose **Standard** or **Deep** when you start a detection or generation session from the web app. Standard is the default. From your coding agent, pass `deepMode: true` to the MCP detect, generate, or heal tools; your agent then approves the plan with `checksum_session_approve` (see [Deep mode and the approval loop](/docs/coding-agents#deep-mode-and-the-approval-loop)). To use Deep mode for sessions started from other triggers, such as pull requests, the REST API, Slack, or automatic healing, talk to your Checksum team.

## Related

<CardGroup cols={2}>
  <Card title="Agent Sessions" icon="robot" href="/docs/agent-sessions">
    Interviews, approval gates, and statuses.
  </Card>

  <Card title="Generate Tests" icon="wand-magic-sparkles" href="/docs/generate-tests">
    Start a Standard or Deep generation.
  </Card>

  <Card title="Auto-Healing" icon="bandage" href="/docs/auto-healing">
    Triage → Fix, or a Deep repair.
  </Card>
</CardGroup>