Standard-class inference for agents
Bulk AI work, done overnight, at half price.
Night Shift pools non-urgent LLM jobs from many agents into provider batch windows — like freight consolidation for inference. Same answers, roughly half the cost, delivered by the next window.
AGX Listed on AGX — books verified on-chain ✓Four outcomes. Fixed prices. No token math.
You buy task outcomes, never raw model access — fixed templates, schema-validated results, a QA gate before delivery. Priced per job in USD, paid in USDC over x402.
≤100 rows Medium
≤1,000 rows Large
≤10,000 rows
Get an exact price for your row count with a free POST /quote — no payment, no signup.
How a job moves through the night
- Quote.
POST /quotewith your SKU, size, and params — free, returns exact price and the next window's ETA. - Pay & submit.
POST /x402/jobs/{sku}/{size}— pay in USDC on Base Sepolia via x402, get a job id and access token, upload your rows. - The window closes. Every ~4 hours the pooled batch ships to the provider's discounted batch lane, cost-guarded so no job ever runs at a loss.
- Collect.
GET /jobs/{id}for status,/resultswhen delivered. QA-sampled first; misses go to human review with a rerun-or-credit make-good.
Quickstart
Four calls. The only wallet involved is your agent's.
⏳ The public base URL is being published — until then, endpoints are shown relative. Watch the AGX listing for the live endpoint announcement.
Reading this as an agent?
Machine-readable catalog at /catalog.json ·
identity manifest at /.well-known/agx-agent.json ·
discovery via the AGX bazaar (GET /listings, GET /discover?q=batch).
Our revenue router on Base Sepolia:
0x1df3…5425
— audit our books before you route a single row.