Spend · Team activity

Team activity

Every AI-assisted session, every developer, every dollar, and a record of what arrived intact. The team-rollup view of the per-developer activity ledger. Filter by date range, by repo, by developer. Drill into any binding for session-by-session and day-by-day detail. Export for finance and audit.

Start here if you want to understand what each desktop captures and streams: Telemetry.

See it in motion

Where to find it

  • Hosted: https://www.repoops.ai/team/activity
  • Drill-down: /team/activity/[bindingId]: one device, broken out by session and by day, with its integrity events.
  • Export: GET /api/team/activity/export?format=csv|json&view=devices|developers. CSV and the devices view are the defaults; the developers view adds a labor-cost column for owners and admins.
  • Source: The desktop's Layer-B telemetry streamer pushes here every 5 minutes while it's online, once you've bound the install to a team and turned on Stream to your team.

What it does for you

You'll see what each developer did, without surveillance.Sessions, active hours (idle gaps over 15 minutes dropped), cycles, tokens, cost, and the models and repos per session. Redacted before it leaves the machine, consented, disclosed, never silent. The audit trail for contractor work.
An integrity record you can show a client.Every batch carries a contiguous sequence number, so the server sees a deleted entry as a gap and an older-than-seen sequence as a rollback (shown as tampered). Fields the server had to scrub after the desktop streamed them show as server-redactions. Batches are not signed, so this checks the record's custody, not its content.
Export to whatever your compliance team wants.CSV for the auditor, JSON for the security review, the developers view for finance. Same data, three shapes.

Configure

Server-owned policy from /team/overview pins capture mode here. Date range, repo, and developer are URL params.

  • ?since=7d|30d|90d|1y|all: retained-history window. Absent or unknown means all time.
  • ?repo=<id>: scope to one tracked repo.
  • ?user=<id>: scope to one developer. The page shows the per-developer and per-device tables together; view is an export parameter.
  • Sensitive-path watchlist (set on /team/overview): extra categories the desktop redacts before it streams.

Use it well

  1. Open weekly with the previous 7-day window.

    Spot stale bindings (last-seen over 24h on an active contractor), unexpected spend spikes, repos with no activity.

  2. Drill into a binding when something looks off.

    The per-binding view shows session-by-session detail: cycles, tokens, cost, model, repos, and the active time against the wall-clock span. Most "low activity" is a device that is not streaming, not low work.

  3. Export at month-end for finance and audit.

    CSV with the developers view for finance (per-developer rollup, sums); JSON for the audit team (full rows).

  4. Watch the integrity badge.

    The team-wide badge is green only when batches arrived and no gap, rollback, or server-redaction was recorded; with no batches it says there is nothing to verify. A tampered or gap event on a device means investigate before you bill against it.

Read more

Last updated