Home/Platform/Activity & Action Receipts
Evidence for live AI work

Trust AI work because
you can inspect it.

Every important live run leaves Activity and a readable Action Receipt. It answers the questions a human operator would ask afterward: what was requested, which SOP it followed, which actions it attempted, who approved, what it did, and what it cost.

hubson · Activity
Activity Runs Receipts
workflow · refund-approvalRun · R-1284
Verify charge in Stripestripe.read_payment · $2,840 · match found
Confirm double-chargezendesk #4821 · two identical charges · 3m apart
Check SOP eligibilityfinance.refunds.v3 · above $2,000 threshold
Issue refund of $2,840awaiting approval · step 4 of 6 · paused 2m agoPAUSED
Refund issuedstripe.refund · approved by [email protected]
Receipt appendedreceipt R-1284 · model usage $0.0042
What a receipt answers

Inspectable by default

A receipt makes the result understandable to a founder, manager, customer, or auditor — no developer needed.

  • What was requested, and which workflow ran
  • Which SOP or context guided the work
  • Which apps/actions were reviewed or attempted
  • What required approval, and who approved it
  • What action happened, and what was skipped or failed
  • What it cost, the outcome, and recommended follow-up
One vocabulary

Activity, Run detail, Action Receipt

Three views of the same work — from the running list, down to the one summary you can share.

ACTIVITYThe listEvery run and its status — running, waiting for approval, done, or failed.
RUN DETAILThe timelineThe full step-by-step execution of a single run, start to finish.
ACTION RECEIPTThe summaryThe readable, shareable record of what happened and what it cost.
hubson · Activity › Run › Receipt
Completedworkflow · refund-approval
Run · Refund approval
Timeline · context → SOP → approval → Stripe refund
Run · R-12846 steps · 1 approval
✓ Action Receipt · R-1284 — readable summary, ready to share
How receipts behave

Readable, transparent, reviewable

Readable by anyone

A founder, manager, customer, or auditor can read it in plain English — no logs to decode.

Complete, not just the happy path

Receipts show what failed, skipped, or was blocked — not only what succeeded.

Model usage is visible

See what each run spent. Bring your own model accounts where supported, or see Hubson-managed model usage separately.

Reviewable later

Review a run later, inspect what happened, and repeat the workflow from the same configuration.

Exportable

Hand a receipt to finance, security, or a customer as evidence of what AI did.

The trust artifact

See what a run hands back

Open a real sample Action Receipt — request, context, SOP, approval, action, cost, and outcome, end to end.