Open-source real-time metering on Kafka with idempotent ingestion, backfills, and entitlement enforcement.
OpenMeter is an open-source usage metering system for engineering teams that need to count product usage accurately at high volume — API calls, tokens, compute seconds — and expose it to billing, dashboards, and entitlement checks in real time. Built on a Kafka-based streaming pipeline, it deduplicates events, aggregates them into meters, handles late-arriving and corrected data, and answers balance queries fast enough to gate a request. It appeals to AI and infrastructure companies that want to own the metering layer rather than couple it to a billing vendor.
Which of the capability map's modules OpenMeter covers — each links to the module's own page, with every tool that supports it.
| Module | Phase | Depth | Note |
|---|---|---|---|
| Fulfill & Bill | |||
| Usage Event Ingestion (API) | Consume & Meter | Core | |
| Streaming Ingestion (S3/Kafka/SFTP) | Consume & Meter | Core | Kafka-native pipeline built for high-throughput event streams. |
| Idempotency & Deduplication Layer | Consume & Meter | Core | |
| Aggregation & Rollups | Consume & Meter | Core | |
| Event Correction & Replay | Consume & Meter | Supported | |
| Late-Arrival Event Handling | Consume & Meter | Supported | |
| Entitlement Management (Feature Flags, Caps, Access) | Fulfill & Activate | Supported | Real-time balance and feature checks enforced from usage data. |
Scored against UsagePricing's Usage-based billing & metering rubric v1.0 (0 weak · 1 adequate · 2 strong), assessed July 2026. Requirements we couldn't verify from public material stay unscored — never guessed. Read the method.
| Requirement | Score | Why |
|---|---|---|
| Real-time balances & drawdown Can a customer (and your product) see an accurate credit or spend balance mid-period? | 2 · Strong | Real-time balances and entitlements are the core product. |
| Correction & re-rating When a meter was wrong, can you fix history without hand-editing invoices? | 1 · Adequate | Reprocessing over the event store within operational bounds. |
| Commits, credits & custom rate cards Can it express how enterprise AI deals are actually signed? | 1 · Adequate | Credits and entitlement grants; negotiated commit contracts are early. |
| Billable-metric flexibility Can finance define a new meter without re-instrumenting the product? | 1 · Adequate | Meter definitions over streamed events with standard aggregations. |
| Invoice & proration correctness Do mid-cycle changes, consolidation, and multi-currency come out right? | 1 · Adequate | Billing features are young next to the metering core. |
| Rev-rec & ERP handoff Can the numbers survive an audit once they leave the billing system? | 0 · Weak | Exports only — finance tooling is out of scope today. |
| Ingestion scale & integrity Does the meter stay correct at production event volumes? | 2 · Strong | Streaming-first, open-source metering built on event infrastructure. |
| Price-change velocity How fast can you ship a pricing change safely? | 1 · Adequate | Plans decoupled from meters; no simulation layer. |
Two things stand out: it is genuinely open source, so the meter — the system of record for what customers consumed — can live in your own infrastructure under your own control, and it treats real-time entitlement enforcement as a first-class use of usage data rather than a billing afterthought. That combination suits teams whose product must check limits on the hot path.
Open-source core plus usage-priced cloud. Self-hosting is free; cloud is priced on events and features.
1 of the companies the Blueprint tracks — from public job posts, engineering blogs, and filings. Every claim links to its evidence on the company page.
No — it is the metering layer underneath one. It produces accurate, deduplicated usage totals and balances; rating those into prices, invoices, and taxes happens in a billing platform or your own code downstream. That separation is deliberate: meters outlive pricing models.
Counting is easy until it has to be correct: duplicate events from retries, late data, backfills after an outage, and real-time balance reads at scale all break the naive approach. A dedicated metering pipeline exists to make usage numbers you can bill on and enforce limits with.
By overlap on the capability map — computed, not curated.