The control plane for autonomous software

Every Run remembered. Only proven code lands.

Verb is a new way to manage code and software work for AI agentic teams. It records work from developers, AI agents, and systems, proves every change, checks how parallel work fits together, and advances the accepted software state only when it is ready.

Open app → See how Verb works
Signal to softwareTrack the full path from incoming work to accepted Base.
Parallel by defaultMany Runs can move at once without losing shared context.
Proof attachedEvidence stays with the exact Run or Set it covers.

Three Missions. Shared payment logic. One coordinated landing.

A real process, not a static dashboard: Missions arrive, Runs start, shared Surfaces are coordinated, Proof is issued, and a proven Set lands.

Mission 1 · Nova
Starting

Update the calculator to apply the new regional tax rule.

    Mission 2 · Atlas
    Queued

    Add settlementReference to the payment output integration.

      Mission 3 · Sage
      Queued

      Add channelCode to inbound payment transactions.

        Verb
        Observing
        Current
        state
        Complete Trail →
        Software development changed

        Source control tracked file versions. Verb manages the full path from incoming work to accepted software.

        Software is no longer produced one change at a time. Developers, AI agents, scanners, and systems can all initiate work. Verb keeps the process understandable from input to output.

        Before
        • Work organized around branches and commits.
        • Context separated from the resulting code.
        • Failed AI Runs disappear from the main workflow.
        • Independent checks do not explain combined behavior.
        • Acceptance depends on a mutable branch flow.
        Now
        • Many Runs can exist for one Mission.
        • Multiple Missions execute at the same time.
        • Signals can come from people, software, or production.
        • Proof is needed for each Run, each Set, and runtime behavior.
        • The full Trail is required for trust, speed, and governance.
        How Verb works

        Signal → Mission → Run → Proof → Set → Land

        Verb terms are useful, but the product has to be understandable as a real operating flow. Each step takes an input, does work, and produces a visible result.

        01

        Signal

        A person, AI agent, or system detects a problem or opportunity.

        02

        Mission

        Verb records the intended outcome, constraints, and Rules for success.

        03

        Run

        One or many Actors perform the work independently and produce an immutable Change.

        04

        Proof

        The result is checked. Pass, fail, or error evidence is recorded and never rewritten.

        05

        Set

        Compatible Runs are checked together so combined behavior is understood before acceptance.

        06

        Land

        A proven Set atomically advances the accepted Base. Nothing is silently discarded.

        StepWhat happensResult
        SignalNew work is detected or requested.Verb opens or proposes work.
        MissionSuccess conditions are recorded.Actors know what to accomplish.
        RunOne or more Actors do the work.A Change is produced.
        ProofThe result is checked.Pass, fail, or error evidence is recorded.
        SetCompatible work is checked together.Combined behavior is understood.
        LandThe accepted state advances atomically.A new Base becomes current.
        Core capabilities

        Built for AI agentic teams that need speed without losing trust.

        Every Run has a Trail

        Preserve the Mission, Actor, context, tools, decisions, Change, Proof, cost, and result even when the Run does not land.

        Proof is part of the work

        Build, tests, security, contracts, approval, and runtime evidence stay attached to the exact Run or Set they cover.

        Parallel work is checked together

        Verb understands Claims, shared Surfaces, and dependencies before acceptance, not only after conflicts appear downstream.

        Autonomy grows by Rules

        Teams decide whether AI agents may observe, propose, prove, land with approval, or operate automatically.

        Trail

        Compact, attributable, and understandable.

        Select any step to see what it means in plain terms. The Trail is the system of record, not a decorative graph.

        Autonomous future

        From developer requests to self-improving software.

        Humans define authority and Rules. Verb makes autonomous action observable, provable, and reversible.

        Continuous loop

        Observe→Signal→Mission→Run→Prove→Land→Observe

        The same system can support a developer asking for a fix today and runtime-directed repair workflows tomorrow.

        Authority ladder

        Each rung is granted deliberately. Nothing escalates on its own.

        Observe Propose Prove Land with approval Land by Rules Operate
        Why Verb

        The future does not need more generated code. It needs a trusted way to decide what becomes software.

        Move faster

        Parallel AI agents can work without losing coordination on the shared parts that matter.

        Trust what lands

        Every accepted Set carries contextual Proof tied to the exact state being advanced.

        Understand everything

        The complete Trail survives beyond the diff, including blocked work, costs, and decisions.

        Build for the autonomous software era

        Join the teams defining how human and AI agent work becomes trusted software.

        Sign in with Google or manual credentials, bootstrap a workspace, and connect this page to a real control plane.

        Open app → Sign in to Verb