Start here
Quick start
Quote a request, then ask it. Two calls.
1. Quote
Price a request before anything runs. No search, no inference, no ledger row.
POST /api/quote
{ "sources": ["exa", "keenable"] }
→ { "total_flop": 0.011, "billed": ["answer", "exa"] }
Add q and model and it also returns the miner's cost, from the FLOP gateway's own zero-spend estimator. That call needs no key and costs nothing, so you can price a whole job — Scout's fee plus the inference behind it — before committing either.
POST /api/quote
{ "sources": ["keenable"], "q": "How many miners are on FLOP?",
"model": "flop:qwen/qwen2.5-0.5b-instruct", "max_tokens": 250 }
→ { "total_flop": 0.001, // Scout's fee
"inference": { "cost_flop": 1.848, // the miner's, gateway-estimated
"input_tokens": 58, "output_tokens": 250,
"gflops": 616, "latency_blocks": 4,
"available_miners": 28 } }
2. Ask
sources: ["auto"] lets Scout pick free sources from the question. A paid source is only ever used if you name it.
POST /api/answer
{ "q": "Which FLOP miners are most reliable?", "sources": ["auto"] }
→ { "answers": [ … ], "pricing": { … } }
Next
See the full endpoint list and pricing.