--- title: "Test the outcome. Let agents choose the path." description: "Agentic testing uses AI agents to write, run, repair, and maintain end-to-end tests. Momentic expands coverage and cuts flaky tests." canonical: "https://momentic.ai/agentic-testing" last-updated: "2026-10-09T17:17:31-07:00" --- # Test the outcome. Let agents choose the path. URL: https://momentic.ai/agentic-testing platform / agentic-testing Describe what should work in plain English. Mix goal-driven AI actions with preset interactions and assertions to check the result. [Try for free](https://app.momentic.ai/signup) [Get a demo](/sales) Works across web, iOS, Android, local development, and CI. ## Describe the action. Assert the outcome. YAML Tests in your repository AI actions Describe a goal; the agent plans and runs the steps. Assertions Check expected behavior with explicit assertions. Step cache Replay cached steps from successful AI actions. CI runs Run saved tests locally and in your CI pipeline. ## How agentic testing works. Define the expected result. Let the agent choose the path, then review the test and its outcome. 01 ### Define the goal Describe one outcome and supply the data it needs. Add preconditions and postconditions to make the expected starting and ending states explicit. 02 ### Let the agent execute the flow The AI action chooses the steps based on the app's current state. Use preset steps when you want to specify each interaction. 03 ### Verify and replay A separate AI reviewer verifies the outcome. Momentic caches successful executions and replays them on later runs. ## Give the action a goal and assert the outcome. An AI action chooses the steps based on the app's current state. Preset steps let you specify individual interactions and assertions. sign-in-verified.test.yaml yaml ```yaml fileType: momentic/test/v2 id: sign-in-verified url: https://www.saucedemo.com/ steps: - act: precondition: A login form is visible. goal: |- Log in with username standard_user and password secret_sauce. postcondition: The product inventory list is shown. - click: The Sauce Labs Backpack product name - assert: The backpack product details are visible ``` Example using the public demo credentials documented in the AI action guide. The postcondition verifies login before the next step opens a product. ### An independent reviewer checks completion A separate AI reviewer checks the app's final state and recorded actions. It verifies whether the action reached your goal. The agent cannot edit or skip your preconditions and postconditions. ### Define one outcome per AI action Give each AI action one goal, such as signing in. Pass test data through variables. Use preset steps when you want to specify each interaction. ### Cache and replay successful executions Momentic caches successful executions and replays the steps on later runs. If replay fails to reach the goal, the agent attempts to repair the flow. ### Separate runtime recovery from a repair Momentic finds changed controls again without editing the test file. Enable Failure recovery to retry eligible preset-step failures in CI. Configure Permanent healing to open a draft pull request with verified test repairs. Read the docs - [How AI actions execute](/docs/core-concepts/agentic-testing) - [Set preconditions and postconditions](/docs/guides/ai-actions/guide) - [Understand recovery and healing](/docs/reliability/auto-maintenance) ## Agentic testing with assertions you control. Cache successful actions and replay them against the current app. Keep explicit assertions to check the expected behavior. ### Give the test a goal Describe one bounded objective and supply the data it needs. The agent chooses the interactions and records the execution trace. Define the result you want to verify. ### Teach agents your product rules Store product terminology, rules for agents, and known flows. Momentic looks up relevant entries for AI-assisted steps; you control the guidance your team adds. Give agents the context they cannot infer. ### Tests from code changes Your coding agent reads the diff with the momentic-spec skill. It uses Momentic to author and run tests for the affected behavior. Review the tests with the code change. ### Keep tests current as the UI changes Momentic finds controls again when the UI changes during a run. Configure Permanent healing to open a draft pull request with verified test repairs. Recover during a run. Review permanent repairs. ### Understand why a test failed Enable classification to attach a category and reasoning to failures. Exit-code overrides are separate, so classification can run without changing your CI gate. Inspect the evidence behind each category. ### MCP authoring and CLI execution Coding agents author tests through the local MCP server. The CLI runs the saved YAML from your configured commit, pull request, deployment, or schedule trigger. Use the same tests before and after review. How engineering teams use Momentic. Read how teams add repository tests, verify changes in CI, and investigate failed runs. “Momentic gave us a fast and reliable way to validate Poe.com's AI responses, even when they weren't deterministic.” Momoko F. Head of Product Operations, Quora 30 min daily test execution, down from 7 hours 500+ manual test cases replaced 100% critical tests created in one month [30 min daily test execution, down from 7 hours "Momentic gave us a fast and reliable way to validate Poe.com's AI responses, even when they weren't deterministic." Momoko F. Head of Product Operations, Quora](/customers/quora) [8x increase in release cadence "It's like giving someone your QA checklist and watching them execute it for you!" Sriram S. Engineering Lead, Source Control, Retool](/customers/retool) [80% faster release cycles "With Momentic, we've caught bugs that would have eluded even our most diligent internal tests." Alex C. CTO, GPTZero](/customers/gptzero) [6x faster end-to-end test creation "We've already seen a 30% decrease in production incidents thanks to Momentic's automated testing." Hanna K. Head of QA, CoverGo](/customers/covergo) [85% reduction in production incidents "Momentic gives us reliable end-to-end coverage, so we can focus on features instead of maintaining tests." Alec H. Staff AI Engineer, Mutiny](/customers/mutiny) [Browse all](/customers) ## Test web, iOS, and Android. Use the same simplified YAML format across web and mobile. Choose the CLI and runtime for each test. [01 Android Run Android APKs on local or hosted emulators.](/android-testing)[02 iOS Native iOS apps and SwiftUI. Same authoring as Android.](/ios-testing)[03 Web End-to-end browser testing on Chromium and Chrome.](/web-app-testing) ## Agentic testing, answered. What is agentic testing? Goal-driven test execution. An AI action chooses the interactions needed to reach a stated outcome. Preset steps and assertions can surround it in the same YAML test. How is agentic testing different from traditional test automation? Preset steps specify each interaction. An AI action chooses the interactions from the live app. The [build vs buy calculator](/comparison/momentic-vs-build-your-own) compares the maintenance costs. How is agentic testing different from AI-assisted testing? An AI action chooses steps toward a goal. A separate AI reviewer verifies the outcome. Authoring assistance can still help create the surrounding preset steps and assertions. Should every test be fully agentic? No. Use preset steps when you need an explicit sequence. Use AI actions when user state, feature flags, or other variation changes the path to the outcome. How does Momentic keep agentic tests trustworthy? A separate AI reviewer verifies the goal. Add preconditions or postconditions to specify the required start and end state. The execution agent cannot edit or skip them. Can Momentic create tests from pull requests? Yes, through your agent. The momentic-spec skill helps a coding agent inspect the diff and write tests for affected behavior. Review those tests in the pull request. Can engineers review and edit agent-created tests? Yes. The tests are simplified YAML files in your repository. Engineers can edit the steps and review an agent's changes with the application code. Can agentic tests run in CI? Yes. AI actions and preset steps run through the CLI. Configure the CI trigger, credentials, runtime, and failure policy before using the result as a gate. Is agentic testing secure? Momentic is SOC 2 Type 2 compliant. Enterprise includes SAML SSO, SCIM provisioning, and audit logs. Evaluate application data and execution location during your security review. How is agentic testing priced? Usage-based credits. AI-generated steps and recovery steps cost two credits rather than one for ordinary steps. Hosted runtime, classification, triage, and AI test selection have separate charges. Still have additional questions? ## Verify what you ship. Run your first test, or book a demo using your app. [Try for free](https://app.momentic.ai/signup) [Get a demo](/sales)