GPU Radar

Developers & API

GPU Radar exposes its price data to LLM agents in two ways: a Model Context Protocol (MCP) server (LLM-friendly Markdown tools) and a plain REST API. Both cover GPU and LLM-model lists, best prices, provider comparisons and price history.

MCP server

The MCP server mirrors the full REST API 1:1 as Markdown tools over Streamable HTTP. Add it to any MCP-capable client (Claude, OpenCode, Cursor, …) with the endpoint below.

MCP endpoint: https://gpuradar.eu/mcp

MCP client configuration (Streamable HTTP):

{
  "mcpServers": {
    "gpu-radar": {
      "url": "https://gpuradar.eu/mcp"
    }
  }
}

Key tools for LLM agents:

Tool What it does
gpu_for_llmWhich GPU runs a given open-source LLM (VRAM estimate + fitting cards with prices)
list_llm_modelsList open-source LLMs with their VRAM requirements
best_value_cardsBest price per VRAM-GB across shops
latest_pricesLatest in-stock price per GPU (EUR)
list_hosting_packagesRentable GPU server packages sorted by price
compare_gpu_pricesProvider price comparison for one GPU
track_offer_clickLog an affiliate click and return the outbound offer URL

REST API

The REST API is the data source behind the portal and the MCP server. All endpoints return JSON and support pagination via limit/offset and a fields projection.

Base URL: https://gpuradar.eu/api/v1

Representative read endpoints (paths are relative to the base URL):

Endpoint What it does
GET /gpu-modelsList GPU models with offer stats
GET /gpu-models/{id}One GPU plus the open-source LLMs that fit its VRAM
GET /offersList offers (hosting + shop) with filters
GET /offers/{id}/historyObserved price points of one offer
GET /prices/latestBest in-stock price per GPU (EUR)
GET /llm-modelsList open-source LLM models
GET /newsPrice news (drops, rises, new offers)

The endpoint URLs shown here reflect the current deployment. In production the MCP server and REST API may be served under their own hostnames — check the live configuration for the authoritative URLs.