# gputape > GPUs you can rent right now, and for how much. Live offers from vast, runpod, shadeform, akash, cudo, prime; refreshed every 5 minutes. Free, no key, no signup. ## For agents - MCP server (Streamable HTTP): https://gputape.com/mcp with tools find_gpus, gpu_prices, price_history - JSON API index: https://gputape.com/v1 ## API - `GET /v1/offers`: Offers. Filters: gpu, min_vram_gb, max_usd_per_gpu_hr, min_gpus, provider, region, tier (datacenter|community), spot=0, available=all, sort=total, limit (default 100). - `GET /v1/gpus`: One row per GPU model: cheapest and median price, availability, providers, 90-day daily median. Optional gpu filter. - `GET /v1/history?gpu=H100&days=30`: Price history for one model (hourly up to 7 days, daily beyond; step=hour|day). Daily points include open/high/low/close of the hourly median. - `GET /v1/status`: When each source last answered. - `POST /mcp`: MCP server (Streamable HTTP). Tools: find_gpus, gpu_prices, price_history. Example: `curl "https://gputape.com/v1/offers?gpu=H100&max_usd_per_gpu_hr=2.5&limit=5"` usd_per_gpu_hr = instance price / GPUs in the instance. Prices are provider asks, not completed rentals.