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

# What is AI-native testing?

> Use AI across the testing lifecycle to build coverage faster, keep tests reliable, and turn run data into product intelligence.

Most test automation uses AI for one isolated task, such as generating a test or
healing a locator. AI-native testing uses agents throughout the development
lifecycle: authoring coverage, choosing and executing tests, maintaining the
suite, and learning from every run.

Momentic connects those stages through tests that remain readable, reviewable
YAML files in your repository. AI reduces the work at each stage without taking
away control. Teams get faster development and more reliable coverage. Run data
explains product risk alongside pass or fail status.

## Use AI throughout the lifecycle

<Steps>
  <Step title="Author">
    Coding agents use your code and a live product session to create or update
    coverage alongside application changes.
  </Step>

  <Step title="Run">
    Runtime agents find elements from their descriptions instead of brittle
    selectors, recover from transient conditions, and select the most relevant
    tests based on a code change.
  </Step>

  <Step title="Maintain">
    Agents classify failures, repair outdated tests, verify changes, and
    quarantine persistent flakes when needed.
  </Step>

  <Step title="Learn">
    Memory, the knowledge base, and the app graph make future decisions more
    consistent while exposing journey coverage and product risk.
  </Step>
</Steps>

## Author with product context

Install the [Momentic skill](https://github.com/momentic-ai/skills) so your
coding agent knows how to add and edit tests. Connect the
[MCP server](/docs/coding-agents/mcp-server) so the agent can combine repository
context with a live browser or mobile session, execute steps, and inspect
artifacts while it works.

For web projects, the [Explore agent](/docs/coding-agents/explore) can bootstrap
coverage for the whole application or identify and cover journeys affected by a
code change. This puts test development in the same loop as product development,
so coverage changes with the product.

<CardGroup cols={2}>
  <Card title="MCP" icon="plug" href="/docs/coding-agents/mcp-server">
    Give coding agents direct access to the product, tests, and run artifacts.
  </Card>

  <Card title="Explore agent" icon="compass" href="/docs/coding-agents/explore">
    Discover user journeys and generate reviewable coverage in your repository.
  </Card>
</CardGroup>

## Run intelligently

Run tests locally, in coding-agent sandboxes, in CI, or on a schedule. Use the
full suite for scheduled runs and bug bashes. For pull requests, AI test
selection uses the code diff and app graph to choose a representative set of web
tests. This shortens feedback without manually maintained dependency rules.

During execution, Momentic uses semantic element resolution and transient
recovery to absorb routine UI variation. Deterministic replay and step caching
keep successful paths fast, while AI handles tasks that require judgment.

<CardGroup cols={2}>
  <Card title="Local runs" icon="terminal" href="/docs/running-tests/running-locally">
    Run web or mobile tests by path, name, or label.
  </Card>

  <Card title="CI/CD" icon="gears" href="/docs/running-tests/ci/custom-setups">
    Configure GitHub Actions, GitLab CI, CircleCI, Jenkins, and other runners.
  </Card>

  <Card title="AI test selection" icon="wand-magic-sparkles" href="/docs/ai/select">
    Use code and journey context to select the most relevant tests based on a
    code change.
  </Card>

  <Card title="Results and reporting" icon="chart-line" href="/docs/running-tests/results">
    Upload runs or generate JUnit, Allure, and JSON reports.
  </Card>
</CardGroup>

See [Performance](/docs/running-tests/performance) for latency benchmarks,
parallelism, and sharding.

## Maintain automatically

When a test fails, Momentic starts with the least invasive response and
escalates only when needed. It can re-resolve a locator, recover from a
transient condition, classify the failure, propose and verify a permanent
repair, or quarantine a persistent flake. This keeps maintenance in the
development loop and leaves durable changes reviewable in Git.

<CardGroup cols={2}>
  <Card title="Locator auto-healing" icon="wand-magic" href="/docs/reliability/auto-maintenance#locator-auto-healing">
    Re-resolve locators and wait for page stability during a run.
  </Card>

  <Card title="Transient recovery" icon="rotate" href="/docs/reliability/auto-maintenance#transient-recovery">
    Clear temporary obstructions without changing the test.
  </Card>

  <Card title="Permanent healing" icon="list-check" href="/docs/reliability/auto-maintenance#permanent-healing">
    Classify failures, repair tests, and deliver reviewable changes.
  </Card>

  <Card title="Quarantine" icon="shield" href="/docs/reliability/auto-maintenance#quarantine">
    Keep flaky tests running without blocking CI.
  </Card>
</CardGroup>

<span id="turn-runs-into-product-intelligence" />

## Turn runs into reusable insights

Each run adds information that Momentic uses to make future AI decisions more
consistent and model how tests exercise the product.

<CardGroup cols={3}>
  <Card title="Memory" icon="brain" href="/docs/ai/memory">
    Reuse relevant decisions from earlier runs so AI behaves consistently.
  </Card>

  <Card title="App graph" icon="diagram-project" href="/docs/ai/app-graph">
    Measure journey coverage, power AI test selection, and analyze product risk.
  </Card>

  <Card title="Knowledge base" icon="book" href="/docs/ai/knowledge-base">
    Give every agent shared product terminology, rules, and known flows.
  </Card>
</CardGroup>

These systems provide reusable context about test execution: coverage across the
product, journeys a change puts at risk, and information the next agent needs
before acting.

See [How Momentic works](/docs/get-started/how-momentic-works) for how preset steps,
AI actions, custom code, and CLI agents fit together.