Worked Yesterday For teams

Local-first · CLI / API / MCP

It worked yesterday.
Here’s what changed.

Run your tests. Capture what changed. Give your agent the evidence—and recheck the result on your terms.

Try the CLI demo Get Worked for your team
Local-first
No telemetry
You approve the run
No automatic repairs
CLI · API · MCP
Your existing tools
worked / changes Example data
FROM04 Sep · 17:42
TO05 Sep · 09:14
4 meaningful changes 3 important
    worked diff --since yesterday

    Try Worked

    Small change.
    Clear evidence.

    From a failing test to a checked result. Try it in a few clicks.

    Scripted demo. No model calls or computer access.

    worked / demo
    1. Test
    2. Evidence
    3. Review
    4. Rerun

    It passed. Then it didn’t.

    worked demo

    checkout-service · unit suite · 12 tests
    Node 22 passes. What changes with Node 24?
    Example details

    A teammate upgrades Node. A test fails. See how Worked connects the two runs without guessing the cause.

    A browser simulation, not a live test or a shell.

    Less manual work

    From a failed test
    to a checked result.

    Compare the environment before a passing run with the one before a failure. Review your agent’s recommendation, then explicitly rerun the approved suite. Evidence, not a guessed root cause.

    • Less setup. Discover existing test commands and approve a focused collection scope.
    • Better context. Runtime versions, dependency changes and separate before/after evidence.
    • Controlled verification. Untrusted recommendations stay separate from recorded test outcomes.

    The complete agent workflow →

    worked / investigate & verify CLI
    1. Run your approved test suite. worked test unit
    2. Use the incident ID printed on failure. worked incident prompt INCIDENT_UUID
    3. After reviewing changes, approve a test rerun. worked incident verify INCIDENT_UUID
    No model required. A passing suite is a recorded outcome, not proof of a complete fix. Team setup

    Worked for your team

    Less guessing.
    More context.

    Bring Worked into your team’s workflow. Contact SVÄRTA AI for commercial access and a package matched to your systems.

    Contact SVÄRTA AI

    CLI 0.10 tested on macOS Apple Silicon. Linux packages are validated per release and target. Platform details

    Does it test my application?

    Worked runs the test command you approve; your existing runner does the testing. Worked adds environment evidence and records explicitly approved reruns. It doesn’t crawl routes, compare screenshots or prove the cause.

    What does it capture?

    New Node setups focus on Git, selected runtimes and approved project files: version requirements, npm locked dependencies and lockfile fingerprints. Inspect your scope or opt into broader system evidence. Missing information stays unknown.

    Where does my data go?

    History stays local. Worked makes no model calls or automatic uploads. Your CI or agent host may share results under its own settings; review masked reports first. Privacy details

    Commercial software by SVÄRTA AI. A separate agreement is required; no general free tier. The demo is not a product license. Earlier license grants and third-party terms remain unaffected. Rights notice.