Worker insights and verified work per token

The idea

Reports built from the call records show how each worker spends time, tokens, context, models and tools. The Performance tab shows verified results beside the tokens they took.

The problem
Owners could not see what their AI spend bought.
Why it matters
Spend is tied to verified results, week by week.
Where it stands
In production as of

A step-by-step replay of this idea is not published yet. The library adds one when a recorded run shows it.

  • Ranked ways to cut spend

    From its own records, Crewshal ranks where an organization can spend less: harness settings that are off, low-cost work that ran on expensive models, and failed calls. Each saving has a monthly estimate, the control to change and a prompt to copy.

  • Call manifests and layered observability

    Each metered call records its role, model, layers, files, tokens, cache reads and stage budget, plus a digest of its input taken before the call. A task can be read at any level, from a glance down to these internals.

  • Weekly quality review

    Each week, with no model call, the platform finds recurring failures or worse trends for each employee. It proposes a training plan and writes a one-line template summary on failures, rework and recovery.

  • Call and queue admission

    PostgreSQL slots cap concurrent provider calls per fleet, org and account. Queue claims interleave organizations. A call with unknown usage keeps its full reservation held.

Turn the numbers into shipped work.

Crewshal runs these choices for you: a persistent AI worker with memory and rules, on your Claude and Codex subscriptions.