BYOC Metering

Metering that runs where your application runs.

1M+ events per second. Inside your VPC. Zero per-event cost.
Same APIs as our managed cloud.

Install with AITalk to a Human

Four reasons cloud metering breaks at enterprise scale.

Throughput ceiling

Cloud metering APIs top out. At millions of events per second, the network breaks before the software does. You need ingestion that scales horizontally, not an endpoint you're rate-limited against.

Latency budget

Enforcement in miliseconds needs colocation. A round-trip to an external cloud adds 50 to 200ms and forces compromises into your request path. Enforcement belongs next to the app, not across the internet.

Data sovereignty

SOC 2. GDPR. FedRAMP. At scale in regulated industries, shipping billions of events to a third-party cloud isn't a preference. It's a compliance blocker. Your data stays in your network.

Cost explosion

Per-event pricing scales with your success. At 1M events per second, that's 86 billion per day, a cost line sometimes larger than inference itself. BYOC removes it as a variable cost.

Kafka. Flink. ClickHouse. Deployed in your VPC.

Not a lightweight agent phoning home. It's the same production infrastructure behind our managed cloud: Kafka for ingestion, Flink for stream processing, ClickHouse for analytics. Deployed as native Kubernetes workloads in your account.

A dual-path architecture built for correctness:

Hot path

Flink runs 1-minute tumbling window aggregations per customer, per feature. Pre-aggregated. Fast to query. Enforcement decisions land in seconds, not minutes.

Cold path

Kafka Connect batch-inserts raw events into ClickHouse. Full historical query surface. Audit trail. Dispute resolution. Every aggregation traces back to its source events.

Not benchmarks. Production specs.

Not aspirational targets. Validated engineering specs.

Metering throughput (BYOC)

1M+
EVENTS/SECOND

Metering throughput (cloud)

100K
EVENTS/SECOND

Event-to-enforcement SLA

99.9%
EVENTS/SECOND

Governance decision latency

Sub-10ms
P99

Metering throughput (BYOC)

Sub-10ms
AT 1,200 REQ/SEC PER INSTANCE
Dimensional aggregation slices usage by any combination of model, feature, region, team, or time window. At ingest time, not query time.

Cost formulas run in the pipeline. Hierarchy enrichment attributes events up the org tree (user, team, department, org) before they hit storage.
Idempotent ingestion means built-in deduplication. No double-counting, even at millions of events per second.

At 1M events/second, per-event pricing costs more than inference.

Cloud metering vendors charge per event ingested. Fine at low volume. Catastrophic at enterprise scale.

At 1M events per second, you generate 86 billion events per day. Even at fractions of a cent each, the metering bill eclipses the inference it's meant to track.

BYOC inverts this. You pay a flat platform fee. The pipeline runs on your compute, at your infrastructure cost. No per-event charge. High-volume usage disappears from your P&L.
Every other vendor charges more as your users grow. We deploy inside your cloud and get out of your cost structure.

Your cloud. Your data. Our engine. Same APIs.

Two planes. One clean trust boundary.

Control plane

Runs in Stigg's cloud
Configuration management
Product catalog
Plan definitions
Entitlement rules
Dashboard
Billing provider integrations
SDK and API management
The brain. Lightweight, secure, maintained by Stigg.

Data plane

Runs in your VPC
Event ingestion pipeline (Kafka, Flink, ClickHouse)
Credit wallet enforcement (TigerBeetle)
Governance decisions (the real-time enforcement engine)
Runs in your Kubernetes cluster. Every event, aggregation, and ledger entry stays in your network.
Same APIs. Same SDKs.
The same API surface as our managed cloud. No code changes to migrate. No vendor-specific abstractions. If your engineers integrated with Stigg, they've already integrated with BYOC.

Deploy one module or all four.

Each module deploys independently into your VPC with the same performance guarantees.

Metering alone

The high-throughput event pipeline. Ingest, aggregate, and attribute usage in your VPC. Feed results to your billing stack, your warehouse, or both. Not a byte leaves your network.

Credits Engine alone

TigerBeetle-backed wallets with real-time balance enforcement. Provably correct. ASC 606-ready. Runs standalone alongside whatever metering you already have.

Governance alone

The sub-10ms decision engine. Evaluates entitlements, budgets, and limits on every request. Runs as a sidecar or a dedicated service next to your app.

The full runtime

Metering feeds Credits. Credits inform Governance. Governance enforces Entitlements. One integrated system in your infrastructure. Same APIs and SDKs as our managed cloud.

Let's talk architecture.

BYOC is an architecture decision, not a checkbox. Book a session and we'll map your event pipeline, identify the right deployment model, and show you exactly what runs in your VPC.

Talk to a Human
Or start with our managed cloud - same engine, same APIs, self-serve signup
AI Credits, Under Control.