
auramatrix
io.github.dadaobear-glitch/auramatrix
LLM API token prices and GPU rental prices, each with its source and date. No key, no account.
Documentation
AuraMatrix MCP server
What AI compute costs, inside your AI assistant: list prices of language-model APIs per channel and GPU rental prices per provider, each with its source and the time it was observed. Coverage: the United States and China.
- Server URL:
https://auramatrix.ai/mcp(Streamable HTTP, remote) - No install, no API key, no account.
- Setup page with one-click buttons: https://auramatrix.ai/connect?ref=github-mcp
This repository holds the connection files only (the server itself is hosted). The data is open under CC BY 4.0 at auramatrix-price-data.
Connect
| Client | How |
|---|---|
| Claude Code | claude mcp add --transport http auramatrix https://auramatrix.ai/mcp |
| Cursor | Add to Cursor, or put {"mcpServers":{"auramatrix":{"url":"https://auramatrix.ai/mcp"}}} in .cursor/mcp.json |
| VS Code | Add to VS Code, or {"servers":{"auramatrix":{"type":"http","url":"https://auramatrix.ai/mcp"}}} in .vscode/mcp.json |
| Claude (web, desktop) | Settings → Connectors → Add custom connector → https://auramatrix.ai/mcp |
| Anything else | Any client that speaks Streamable HTTP: give it the URL above |
Ask it
- Where is Claude Opus 5.5 cheapest? Give me the source.
- I send 20M input and 4M output tokens a day. What does that cost on three flagship models?
- What does an H100 cost on demand at each provider right now?
- Which models changed price this week?
- I was quoted $2.10 per hour for H100s on a one-year contract. Is that good?
Tools
Ten tools only read. One, share_price, stores a price, and only after the user explicitly agrees.
list_models: models with official input and output price and the lowest-priced channelget_price: one model on every channel, with region, plans, trust status and sourcecompare_channels: channels for one model, sorted by price, with the difference from the official pricecheapest: models ordered by their lowest input priceprice_changes: observed and vendor-announced price changesestimate_cost: cost of one request from token counts and the channel's billing rulesgpu_prices: GPU rental prices per provider (on demand, spot, reserved, marketplace)price_index: token price indices by tier and GPU-hour reference prices (trial)measured_performance: measured time to first token, output speed, answer rate, charge against list pricecheck_price: is a token price or a GPU quote the user was given a good one, against the prices on recordshare_price: adds a price the user pays, anonymously and only with their explicit consent, to the benchmark of what buyers pay
What the numbers are
List prices as published, not invoices. Every answer names the channel or provider, the observation time and the source page. Currencies are never converted. A missing price is unknown, not free. How the data is collected and checked: https://auramatrix.ai/prices/methodology?ref=github-mcp. Privacy: https://auramatrix.ai/account-data?ref=github-mcp.
Contact: hello@auramatrix.ai
No installation packages available.