Tracked spend, evidence, and coverage gaps
This lane now shows real run-ledger spend and says plainly where the rest of the cost model is still missing.
Real AI-run spend is visible now, but it is intentionally no longer pretending to be the whole budget.
The correction path starts by showing what the ledger really knows, how fresh it is, and which cost lanes still need adapters before they deserve numbers.
AI run spend is real. Hosting, SaaS, and vendor invoices are still pending.
Tracked spend composition
Canonical run-ledger spend for the last 30 days. This is real, but it does not include hosting or SaaS invoices yet.
If this gets stale, the cockpit should say so plainly instead of implying live cost coverage.
Most tracked spend is concentrated in one model family, which is a useful guardrail signal on its own.
Spend by agent
What deserves review
Current ledger freshness is 23d old, so this page must stay explicit about recency.
product.idea.intake.coach accounts for 95% of the ledger-backed cost window.
VPS, Cloudflare, analytics, and tooling invoices should not be shown as numbers until their real sources are wired.
Most expensive recorded runs
product.idea.intake.coach · $0.05 · gpt-5.2-2025-12-11 · 21 Feb
product.idea.intake.coach · $0.05 · gpt-5.2-2025-12-11 · 21 Feb
product.idea.intake.coach · $0.05 · gpt-5.2-2025-12-11 · 21 Feb
product.idea.intake.coach · $0.05 · gpt-5.2-2025-12-11 · 24 Feb
product.idea.intake.coach · $0.05 · gpt-5.2-2025-12-11 · 21 Feb