# Modelmeter > Agent-first measurement and instrumentation for AI models. JSON-first APIs, stable URL contracts, no authentication required for V1. Modelmeter exists so AI agents can answer questions like "what does this LLM call cost?" or "will this fit in this model's context window?" without scraping a JS-heavy SPA. Every tool is reachable via a plain GET URL and returns JSON. Pricing data is community-maintained in a public GitHub repo; each entry carries a `last_verified` date so callers can judge freshness for themselves. ## Stable endpoints - [/estimate](https://modelmeter.xyz/estimate?model=anthropic/claude-sonnet-4-6&input=1000&output=500): cost of an LLM call given model, input tokens, output tokens. Includes upstream markup comparison for reseller entries. - [/models](https://modelmeter.xyz/models): list of available models with current pricing. Filters: `provider`, `include_unverified`, `include_unavailable`, `exclude_tags`. - [/events](https://modelmeter.xyz/events): AI market events (model launches, deprecations, regulatory actions, compute partnerships, etc.) for annotating the pricing timeline. Filters: `provider`, `type`, `model`, `since`, `until`, `include_unverified`, `limit`. - [/pricing.json](https://modelmeter.xyz/pricing.json): raw pricing snapshot, identical to `pricing/current.json` in the repo - [/events.json](https://modelmeter.xyz/events.json): raw events snapshot, identical to `events/current.json` in the repo - [/openapi.yaml](https://modelmeter.xyz/openapi.yaml): OpenAPI 3.1 spec for the API ## Source of truth - Pricing data lives at https://github.com/modelmeters/modelmeter - Daily date-stamped snapshots are committed to `pricing/snapshots/` so the historical dataset accumulates - Found a price change? Open a PR ## Limits V1 is unauthenticated but rate-limited per IP. An API-key tier with higher limits is planned for agents under sustained load. ## Roadmap - Context-window-fit checker - Embedding cost estimator - Real-world latency aggregator - Tool-calling schema validator