[AGENT API]

The AI compute intelligence API for agents.

A JSON + llms.txt interface to the Servers.Computer index. Built for autonomous agents, RAG pipelines, procurement bots, and inference routers that need real GPU pricing — not stale screenshots.

Endpoints

GET/api/gpus

List all indexed GPUs with TFLOPS, VRAM, $/hour, and provider availability.

GET/api/gpus/{slug}

Full spec sheet for a single GPU (H100, H200, B200, MI300X, etc.).

GET/api/providers

All clouds & neoclouds with on-demand and reserved pricing tiers.

GET/api/benchmarks

Throughput, latency, and $/TFLOP benchmarks normalized across providers.

POST/api/compare

Compare 2–5 GPUs or providers. Returns a structured diff for agent consumption.

POST/api/quote

Get a quote for N GPUs × hours across providers, sorted by total cost.

GET/llms.txt

Agent-readable site map (LLM-citation format) with canonical URLs and summaries.

Examples

Find the cheapest H100 right now

curl https://servers.computer/api/gpus/h100?sort=price_asc&limit=3

Compare H200 vs B200 for inference

curl -X POST https://servers.computer/api/compare \
  -H "Content-Type: application/json" \
  -d '{"gpus":["h200","b200"],"workload":"inference"}'

Quote 64×H100 for 720 hours

curl -X POST https://servers.computer/api/quote \
  -d '{"gpu":"h100","count":64,"hours":720}'

Authentication

Public read endpoints are free and rate-limited to 60 req/min per IP. Higher tiers, write endpoints, and webhook subscriptions require an API key.

→ Get an API key → Full docs