What it does
The server connects MCP-compatible assistants such as Claude, ChatGPT, Cursor, Windsurf and VS Code to price data on AI inference. It exposes 23 price indexes across modality, channel, tier, license, origin and use case, each with a benchmark level and a spot price in dollars, along with nine market KPIs and six model capability measures.
It suits questions such as where a model is cheapest to run, how open-weight pricing compares with proprietary pricing, or how prices have moved since May 2026. Most tools are free. Vendor names, model names and exact prices are available on the PRO tier.
For the hosted server, add it as a custom connector in Claude, or through a config file in clients that accept a remote URL; clients that do not use a proxy. PRO users pass their key as _atom_api_key on each call. The code is open for self-hosting and needs a Supabase database URL, plus a server-side service key for hosted use. It is built with TypeScript and Node.js and runs over stdio or HTTP.
Limitations and requirements
Vendor names, model names and exact prices require a PRO subscription, listed at $500 per month, and a key passed with each call. Self-hosting requires a Supabase database URL, and the hosted setup requires a server-side Supabase service role key. Clients without remote URL support need the proxy. The publisher notes that spot changes reflect models arriving and leaving as well as repricing.
Safety
Our team has not reviewed this resource. Only automated rules checked it; resources with high-risk signals are not published. These are signals, not a guarantee that installing it is safe. Try it in a sandbox project first.
- Runs code or changes an assistant's behaviour
Sources
- Official MCP Registry first seen 13 Nov 2026 · last seen 13 Nov 2026 · its page on the source
Attic Standard
The global price benchmark for AI inference: 23 price indexes, market KPIs and SKU-level prices.
v2.1.0As shown on registry.modelcontextprotocol.io when we last read it