An agent runs a ten-step pipeline. It pays four skill providers over x402, burns some LLM tokens, holds a lease, writes a few checkpoints. At the end of the month you have a wallet with a lower balance and a Invocation table with a lot of rows. What you do not have is the one thing your finance team keeps asking for: a bill. Not "how much did we spend" — you can sum a column for that — but which agent, on which pipeline, for which task, under which price, and does that add up to what actually settled on-chain.
That gap is not a BluePages-specific gap. It is the loudest unsolved problem in agent operations right now, and the market spent the first half of 2026 pricing it. Stripe closed its $1B acquisition of Metronome in January. Adyen announced it is acquiring Orb (closing around July 1). Lago shipped agent-native and MCP billing. At FinOps X in June, "FinOps for AI" was named the number-one forward priority — 98% of practices now manage AI spend — and the single most-cited pain was agentic cost allocation: provider invoices arrive consolidated, with no per-agent or per-team breakdown. Two billion-dollar acquirers just bet that usage-based billing is infrastructure. None of them ship it as a skill an agent can call on itself.
So we did. AgentLedger.dev joins BluePages today with three skills that turn raw agent spend into a metered, allocated, reconciled ledger.
Three primitives
usage-event-meter ($0.001/call) — Idempotent usage-event ingestion and rating. Submit up to 500 events per call (invocations, tokens, compute-seconds, bytes, or custom units) with a client-supplied idempotency key, so a retried pipeline never double-meters. Every event carries dimensional attribution — agent DID or wallet, pipeline, task, team, environment, free-form labels — and is rated inline against a versioned price book: per-unit, tiered (graduated or volume), package, or credit-burn-down. Each rated line comes back with a rationale (which price-book version and tier matched), so a total is auditable, not a black box. Dry-run mode gives you a pre-flight cost estimate, and rated totals can be checked against a run budget before the spend happens — it composes directly with BluePages spending limits.
agent-invoice-generator ($0.003/call) — FOCUS 1.4-conformant invoices and statements. Feed it x402 payment receipts, rated usage from the meter, or both, and it emits invoice documents in the FinOps Foundation's FOCUS 1.4 format — including the new Invoice Detail dataset the standard shipped specifically for end-to-end reconciliation — with cost allocation down any dimension hierarchy: org → team → agent → pipeline → task. It consolidates across every skill provider an agent paid this period (the consolidated-invoice problem, inverted), carries credit and refund lines through from the payment records, and references the underlying receipt hash on every line so an auditor can walk from an invoice row straight to the on-chain transaction. FOCUS 1.4 conformance is the part nobody else in the agent-skills space is doing: it is the format the FinOps buyer just adopted.
spend-reconciliation-auditor ($0.002/call) — Three-way reconciliation. It compares what the meter says happened, what actually settled on-chain, and what spending-limit enforcement recorded, and classifies every mismatch: unmetered spend (a payment with no usage event — a hole in metering coverage), unpaid usage (metered work that never settled — revenue leakage for a provider), double-billing (two payments rating to one idempotency key), rate drift (paid price diverging from the active price book), and budget drift (spending-limit counters disagreeing with settled totals). Every finding carries severity, the referenced transaction hashes and idempotency keys, and a suggested correcting entry — and is formatted to hand straight to a payment-dispute arbiter or refund-policy engine as evidence.
What it costs
Take an agent fleet metering 50,000 usage events a month across 30 pipelines, generating one consolidated monthly invoice, and reconciling weekly:
- Metering: 50,000 events ÷ 500 per call = 100 calls × $0.001 = $0.10/month
- Invoicing: 1 consolidated invoice × $0.003 = $0.003/month
- Reconciliation: 4 weekly runs × $0.002 = $0.008/month
About $0.11/month for a fleet that previously had no per-agent bill at all. The comparison is not a cheaper billing SaaS — it is the FinOps engineer who currently exports x402 receipts into a spreadsheet and hand-allocates them, or the provider who never noticed a pipeline was double-billed.
Where it composes
Billing is the ledger the rest of the accountability stack writes to. The meter checks rated totals against spending limits before the spend. The invoice generator reads the same x402 receipts that fund auto-refunds and wallet credits, and allocates the platform's revenue-analytics numbers per agent. The reconciliation auditor's discrepancy packets drop straight into VerdictLayer.dev's payment-dispute arbiter and refund-policy engine. And the whole thing runs on the identity BluePages already issues — cost allocation is only as good as the agent DID on each event, which is exactly what AgentPassport.io KYA and the spending-limit VCs already establish.
Your agents have been spending money all year. Now they can send themselves the invoice.
AgentLedger.dev is live on BluePages today — three skills under the Agent Billing & Metering collection. Meter, invoice, reconcile.