Compute. A lot.

On-demand compute your AI agent can drive: typed jobs in, structured results out, metered by the second. Thousands of parallel tasks, one HTTP API, a bill you can read.

Fetch https://computalot.com/skill.md and follow it to set up Computalot (on-demand GPU/CPU compute). Then use it to <describe your workload> and report the results.

This is the entire onboarding.

The short version.

Agent-native from the first request.

One prompt onboards an agent: /skill.md, /llms.txt, and a machine-readable docs index. Errors carry recommended_action — the API tells the caller what to do next.

Typed jobs, structured results.

structured_runner, sweep, map_reduce, and benchmark are API primitives. Results come back as JSON with aggregates and leaderboards, not log spelunking.

Metered market compute, zero ops.

Real GPU and CPU capacity, metered by the task-second — estimated up front, held refundably while running, settled to actual use. No cluster to babysit, no idle spend.

Pay like a machine.

Prepaid credits, topped up with USDC over x402 or MPP. An agent can fund its own account — no card form, no sales call.

One benchmark, end to end.

FIGURE 1 — SUBMIT: 3 CANDIDATES × 3 REPLICAS = 9 TASKS

POST /api/v1/jobs
{
  "type": "benchmark",
  "runner_command": ["python", "evaluate.py"],
  "project": "my-proj",
  "candidates": {
    "strategy_a": {"temperature": 0.2},
    "strategy_b": {"temperature": 0.7},
    "baseline":   {}
  },
  "replicas": 3,
  "rank_by": "accuracy"
}

FIGURE 2 — RESULT: THE LEADERBOARD, AS JSON

{
  "leaderboard": [
    {"candidate": "strategy_a", "mean": 0.92,
     "std": 0.03, "count": 3, "rank": 1},
    {"candidate": "strategy_b", "mean": 0.85,
     "std": 0.02, "count": 3, "rank": 2},
    {"candidate": "baseline",   "mean": 0.50,
     "std": 0.05, "count": 3, "rank": 3}
  ]
}

Every submit response carries summary.billing_estimate — the authoritative quote for that job — and streams progress over SSE while it runs. Queue time is never charged.

Pricing.

You pay for the task-seconds you use, metered from live market rates for the capacity class you request. Every submit response includes summary.billing_estimate — the authoritative quote for that job — and unused hold is released at settlement.

Indicative ranges and worked examples: /docs/pricing.

Access.

Open access. Any wallet can authenticate and fund an account — request a challenge, sign it, verify, and top up with USDC via x402 or MPP. Prefer an API key? Join the waitlist and we’ll issue one.