Atlas

    Atlas records what coding agents actually did.

    Every governed run leaves a readable record. Atlas keeps it, so you can inspect what happened instead of trusting an agent's "done".

    What you can inspect

    Outcomes

    Whether the run finished, stopped, or needs review.

    Verification

    Whether the configured verifier actually passed.

    Failures

    Which of the 13 failure classes appeared, and when.

    Retries

    How many attempts the run took before it settled.

    Receipt evidence

    Whether the run receipt checks out cryptographically — reported separately from task verification.

    Cost evidence

    Spend where it was recorded. Unknown cost is shown as unknown, never as $0.

    SANSA analyzes patterns across those records — recurring failures, retries and wasted spend.

    Public research views