What it does
Model Meter answers: 'Which project, feature, or team is spending the most on AI, and how does that compare to last week?' It tracks every LLM API request across providers, attributes costs to specific projects or features, and displays spend trends on a live dashboard — all with a two-line SDK integration.
- Every LLM call, across every provider, captured automatically via OpenTelemetry
- Know exactly which feature, service, or team is spending what
- Spend trends updated in near real-time — not end-of-month
- See cost vs. quality side-by-side before switching models in production
- Match your request-level estimates against actual provider bills
- Prompts and completions are never collected — cost metadata only
What it does
Model Meter is a TypeScript SDK, ingestion API, cost engine, and live dashboard — one system for tracking and attributing AI spend.
- TypeScript SDK (@model-meter/sdk) — init() once, all AI calls tracked
- Auto-tracking via OpenTelemetry gen_ai.* spans — zero code changes after init()
- Batch ingestion API with per-event validation and idempotency
- Decimal-precision cost engine with effective-dated pricing
- React dashboard — spend trends, provider/model/project breakdown
- Multi-workspace data model with tenant isolation
- Auth, workspace UI, and key management
- Hosted cloud — managed dashboard
- Gateway adapters (LiteLLM, Portkey)
- Workspace budgets and threshold alerts
- Provider billing import and reconciliation
Who it's for
Best fit: B2B SaaS and AI-native teams with 5–50 engineers using multiple LLM providers who need to attribute AI costs to specific features or projects and want to stop surprise AI bills before they arrive on the invoice.
- Engineering managers who need per-project cost breakdowns
- Backend engineers who want visibility without changing existing code
- Finance and FinOps teams reconciling AI costs with provider invoices
- AI/ML engineers comparing cost-per-request across candidate models
- Startup founders who need observability without another SaaS bill
Not designed for
Model Meter is not trying to be everything. It is a focused cost visibility and spend attribution tool.
- Tracking personal ChatGPT subscription usage
- Teams requiring enterprise SSO or SOC 2 from day one
- Replacing full-featured LLM observability platforms (LangSmith, Langfuse)
- Universal LLM proxy with caching
Built in the open
The Model Meter source code is at https://github.com/jayanti-prajapati/model-meter — inspect it or contribute pricing fixtures, gateway adapters, dashboard improvements, and documentation. Hosted plans will be paid; the source stays available.
