Cost & Receipts

Measure cost-per-success, not just tokens — and get a shareable receipt for every task.

BudgetAI measures what a task actually costs to complete and issues a receipt for every run. Savings become something you can show, not just claim.

Cost-per-success

Token counts alone hide the real question: how much did it cost to actually finish the job? BudgetAI tracks three headline numbers:

MetricMeaning
SavedSpend avoided versus always using the top tier
Success rateShare of tasks that met their contract
Cost per successAverage cost of a completed task

Receipts

Every session ends with a receipt showing where the money went and why any escalation happened.

receipt · session 4f2a                    PASSED
────────────────────────────────────────────────
router · R1 (verify)                        $0.012
escalation · R2                             $0.061
tools · 5 calls                             $0.007
────────────────────────────────────────────────
total                                       $0.080
saved vs. top tier                          $0.410

Export a receipt to share with a teammate or client:

budgetai cost receipt <session-id> --format md   # or pdf

Inspecting spend

budgetai cost                 # recent spend summary
budgetai cost --session <id>  # one session in detail
budgetai cost --by provider   # group spend by provider

Shadow rehearsal

Estimate a run before committing to it:

budgetai agent -m "Refactor the parser" --dry-run
# → estimated $0.12–$0.40, success ~80% under this policy

Pair receipts with Channel Budgets to keep every surface — personal chat, team workspace, and cron — inside its own spend ceiling.