Unimeter

The open-source usage metering engine for teams that bill by what customers actually use.

Unimeter counts, aggregates, and queries usage events in real time. One binary, no external infrastructure, deploys anywhere from a $5 VPS to a multi-node cluster.

Why Unimeter

Most usage metering setups require a message broker, a database, and a batch pipeline glued together with custom code. Unimeter replaces all of that with a single self-contained server:

• Real-time queries — ask for a customer's current usage and get an answer in under a millisecond.
• Exactly-once — duplicate events are detected and discarded automatically. Your totals stay correct even when clients retry.
• Replication — 3-node clusters with automatic leader election. Kill a node and the cluster keeps going.
• Zero dependencies — no external databases, no message queues, no third-party libraries. One binary, nothing to install alongside it.

Quickstart

docker run -d --name unimeter \
  --security-opt seccomp:unconfined \
  -p 7001:7001 -p 9090:9090 \
  ghcr.io/unimeter/unimeter:latest

Install the Go or Python SDK, then:

client, _ := billing.New([]string{"localhost:7001"})

// Define a metric
client.Metrics.Create(ctx, billing.MetricSchema{
    Code:            "api_calls",
    AggType:         billing.AggCount,
    PeriodType:      billing.PeriodCalendar,
    BillingCycleDay: 1,
})

// Record usage
client.Ingest(ctx, []billing.Event{
    {AccountID: 42, MetricCode: "api_calls", Value: 1},
})

// Query
usage, _ := client.Query(ctx, billing.QueryRequest{
    AccountID: 42, MetricCode: "api_calls",
    Period: billing.CurrentMonth(),
})
fmt.Println(usage.Value.Count) // 1

See the full quickstart for a step-by-step walkthrough, or browse the examples for complete runnable scenarios in Go and Python.

Features

Aggregation types

Type	Description
COUNT	Number of events in a period
SUM	Sum of event values
MAX	Largest value observed
LATEST	Most recent value by timestamp
COUNT UNIQUE	Distinct values with add/remove operations

Billing periods

• Fixed windows — configurable duration (default 30 days).
• Calendar months — variable-length months with optional billing cycle start day (1–28).

Dimension filters

Attach properties to events and query usage sliced by any dimension. Supports single-dimension queries (provider=aws) and multi-dimension AND queries (provider=aws AND region=us-east).

Alerts

Up to 8 thresholds per metric, checked on every event. Alert history accessible via query API and push subscriptions in the SDK.

Replication

3-node clusters with automatic leader election. If a node goes down, the remaining two continue serving reads and writes. When it comes back, it catches up automatically.

Observability

Built-in Prometheus metrics endpoint with a Grafana dashboard and k6 load test scenarios.

SDKs

Language	Repository	Install
Go	unimeter/go-unimeter	go get github.com/unimeter/go-unimeter
Python	unimeter/python-unimeter	pip install unimeter-python

Building from source

Requires Zig 0.16.0.

zig build              # build
zig build test         # run tests
zig build run          # start a single node

Or use Docker:

docker build -t unimeter .
docker run --rm unimeter

Documentation

Full documentation at unimeter.io:

• What is Unimeter
• Quickstart
• Concepts
• Go SDK
• Python SDK
• Cluster operations
• Stripe integration guide
• Benchmarks

Examples

Complete, runnable examples in Go and Python: unimeter/examples

Example	Description
saas-api	API call counting and monthly usage query
seat-based	COUNT_UNIQUE seats with add/remove
infra-metering	Dimension filters, per-provider breakdown
high-throughput	Buffered async ingest at high event rates
free-tier-alerts	Alert thresholds and enforcement
stripe-integration	Stripe webhook → usage → invoice

License

O'SaaSy — MIT with SaaS restriction. Free to use, modify, and self-host. Commercial SaaS offering of the software itself requires a separate agreement.