Agent-native access · same data, zero scraping
The registry, inside your assistant
Every question this site answers — current prices, fair-price bands, provider reliability, model fit, $/M-token math — is available as MCP tools. Ask Claude "what's a fair price for an H100 cluster right now" and it can answer from this week's graded registry, with the citation attached.
Connect
claude mcp add --transport http compute-assay https://computeassay.com/api/mcp
Or any MCP client (Cursor, VS Code, custom agents): Streamable HTTP endpoint at https://computeassay.com/api/mcp. No auth, no key, no rate-limit games — same CC BY 4.0 terms as the open data: attribute Compute Assay, nothing else.
Tools
| Tool | What it answers |
|---|---|
list_accelerators | What's tracked, seller counts, price ranges |
gpu_prices | Every listing for a GPU: price, grade, fabric, terms, source |
fair_price_band | The negotiation anchor: Grade A/B walk-up band + cheapest committed rate |
provider_record | One provider: offerings, published SLA/terms, what they don't publish, 90-day incident history |
model_fit | Cheapest GPUs that fit N-billion params at a precision |
cost_per_million_tokens | Rented $/M output tokens from sourced benchmarks, assumptions stated |
Why this exists
Assistants are how most of these questions get asked now. An assistant browsing the web gets whatever a search index cached; an assistant with this endpoint gets this week's graded registry with sources. Every response carries a citation field — if you build on it, keep the attribution and the snapshot date. Grades describe what sellers publish, not measured performance (methodology).