MCP server
Compute prices, inside your editor
A free MCP server that answers "what's the cheapest H100 right now?" and "who serves DeepSeek V4 cheapest?" directly in Claude Code, Cursor, or Claude Desktop. Same data as this site, refreshed daily. No API key, no signup.
Data refreshed 2026-07-09 · endpoint https://llmhosting.ai/mcpInstall
Listed in the official MCP registry as ai.llmhosting/pricing — clients that browse the registry can install it by name.
Claude Code — one line:
claude mcp add llmhosting --transport http --url https://llmhosting.ai/mcpCursor / Claude Desktop / any MCP client — add to your MCP config:
{
"mcpServers": {
"llmhosting": { "type": "http", "url": "https://llmhosting.ai/mcp" }
}
}Tools
| Tool | What it answers |
|---|---|
| search_models | "Which DeepSeek models have API pricing?" |
| get_model_prices | "Every provider price for Llama 3.3 70B, cheapest first" |
| list_gpus | "All GPUs with live rental pricing" |
| cheapest_gpu | "Cheapest GPU with at least 80 GB VRAM right now" |
| gpu_price_history | "How has the H100 floor moved over 30 days?" |
Try these prompts
- "What's the cheapest way to run DeepSeek V4 Flash via API right now?"
- "Compare H100 vs H200 rental floors and tell me if the premium is worth it for a 70B model."
- "Has RTX 4090 marketplace pricing dropped this month?"
Prices are USD — per 1M tokens for models, per GPU-hour for rentals. The server is stateless and free for any use; data terms match the free API(CC BY 4.0, link back appreciated). Heavy programmatic use should hit the JSON endpoints directly.