Skip to content
runward
runward
Product
Who it is for
Trust
Documentation
News
Try it on an example Talk to us

The runward engine

How runward works, in about a minute.

  1. 01Your agents build
    AI
    AI“it’s done”
    AI

    Your AI agents build fast.Faster than anyone can review it all.So when one of them says “it’s done”…who checks?

  2. 02Your decisions, written
    ADR-000112 May 2026accepted
    ADR-000212 May 2026accepted
    ADR-000312 May 2026accepted
    acceptedAI

    The team’s record

    With runward, everything starts from your decisions.Your agents draft most of them.You make the call.

  3. 03Each rule, its evidence
    ADR-0001AIADR-0002AIADR-0003AI

    1 Frame

    AI

    2 Architect

    AI
    model-provider.port.tsTriage​Model​Port

    3 Floor

    AI
    guard.tsguardFields

    4 Iterate

    5 Govern

    AI
    guard.tsguardFields

    6 Hand over

    AI
    triage.test.ts

    Those decisions are given to the agents before they build.And every rule they apply has to say where its evidence is, in the code.

  4. 04Six steps, without AI

    1 Frame

    2 Architect

    model-provider.port.tsTriage​Model​Port

    3 Floor

    guard.tsguardFields

    4 Iterate

    5 Govern

    guard.tsguardFields

    6 Hand over

    triage.test.ts

    At every delivery, runward reads the record again.Without AI, step by step.Does each piece of evidence still point to the code?

  5. 05A name changes
    Tests: passed

    1 Frame

    2 Architect

    model-provider.port.tsTriage​Model​Port

    3 Floor

    guardFields
    AIguard.tsguardFieldsapply​Field​Guards

    4 Iterate

    5 Govern

    guardFields
    AIguard.tsguardFieldsapply​Field​Guards

    6 Hand over

    triage.test.ts

    Here’s a real case.An agent renames a function.The tests stay green.But two pieces of evidence cite a name that no longer exists.

  6. 06Refused, and where
    Reading of the check

    runward-demo-renaming

    Refused by the check
    branch of the change · commit 97fce96 · check of 10 Nov, 15:52
    1. 1. Frame2 documents
    2. 2. Architect5 documents
    3. 3. Floor1 refusal
    4. 4. Iteratenothing required
    5. 5. Govern1 refusal
    6. 6. Hand over2 documents
    First-version measurements: a cited piece of evidence is refused by the check.runward/floor.md · line 22
    Security assessment: a cited piece of evidence is refused by the check.runward/governance/threat-model.md · line 50

    runward refuses.And it says where: the file, the line.The check stays red.

  7. 07Two lines
    Nothing is blockingaccording to the team’s requirements · branch of the repair · commit 0b56196

    1 Frame

    2 Architect

    model-provider.port.tsTriage​Model​Port

    3 Floor

    guardFieldsapplyFieldGuards
    guard.tsapply​Field​Guards

    4 Iterate

    5 Govern

    guardFieldsapplyFieldGuards
    guard.tsapply​Field​Guards

    6 Hand over

    triage.test.ts

    The agent fixes two lines.Nothing blocks any more.Going to production is still your decision.

  8. 08The same verdict
    The team’s delivery chain
    Nothing is blocking

    commit 0b56196 · runward 0.46.0

    Your security team
    Nothing is blocking
    • Same code0b56196
    • Same versionrunward 0.46.0
    • Same resultNothing is blocking

    And you don’t have to take our word for that verdict.Your security team can rerun it themselves, and get the same one.Same code, same version, same result.runward is open source.You can use it with your own agents.

0:00 / 1:18

Talk to us

Tell us who you are, and we will get back to you.

Your details are recorded by Another Framework, publisher of runward, to get back to you. They are neither sold nor shared, and are deleted three years after our last exchange. To see or delete them: contact@runward.dev.