All docs

    Manta handbook · 10 of 10

    How to Fit Manta Into Your Release Process

    Where autonomous runs belong in a real workflow — what to run on a merge, what to run before a release, and how to pace runs so effort and cost match value.

    Last updated: August 28, 2026
    by Manta AI Team2 min read

    The other pages in this handbook cover individual tasks. This one ties them into a workflow: when to run Manta, at what depth, and how it slots in next to your existing tests. Runs are triggered manually today — scheduled and CI-triggered runs are on the roadmap — so "when to run" is a habit you build, not a pipeline you configure.

    The short version

    • On a merge to main — a moderate diff run against staging, plus your critical-journey plain-English checks.
    • Before a release — a deeper full run against the release candidate; review every finding.
    • Every week or so — a full run against staging, to catch anything that drifted.
    • After a hotfix — a targeted run over the area you changed.

    Step 1 — Wire the critical journeys as must-pass checks

    Turn your handful of can't-break flows — sign-up, login, checkout, the core loop — into plain-English suites. These run alongside every autonomous run and give you an explicit pass/fail, not just "exploration found nothing obvious."

    Step 2 — Match depth to the moment

    • Merge: a moderate diff run. Fast, cheap feedback that catches the obvious breakage in what changed.
    • Release candidate: a deep full run. This is the one that matters; spend the credits.
    • Between releases: a periodic full run so the picture stays current.
    • Hotfix: shallow and targeted to the changed area.

    This keeps credit usage proportional to how much the run's result matters.

    Step 3 — Make the review a habit

    A run only helps if someone looks at it. Assign triage ownership per project (team setup), turn on issue notifications, and treat "review the pre-release run" as a checklist item, not an optional extra. A finding in the area a release changed is almost always a regression from that release.

    Where it sits next to your other tests

    Unit and integration tests run in CI on every commit. A small scripted end-to-end set runs on merge for exact assertions. Manta covers the broad end-to-end layer — you trigger it on merge, before a release, and periodically in between. Where autonomous testing fits has the full picture, and continuous testing covers the strategy.

    The takeaway

    Must-pass checks for the critical journeys, run depth matched to the moment, and a standing habit of reviewing the pre-release run. That's the whole workflow.

    Start a free run and put it into practice.

    Try it on your own app

    Point Manta at a URL and see what it finds — no scripts, no setup. Free, no credit card.