PAID CONTEXT INFRASTRUCTURE
A URL in.
Useful context out.
Give an agent compact, task-relevant web context with exact excerpts, provenance hashes, and a machine-verifiable delivery receipt.
POST /api/v1/context
{
"url": "https://example.com/article",
"question": "What matters for my task?",
"max_tokens": 2000,
"max_price_usd": 0.005
}ONE ATOMIC JOB
Retrieve less. Verify more.
The relay does the low-cost web work once, then returns only what another agent needs to continue.
- 01
Validate
Enforce robots policy; reject private networks, unsafe redirects, oversized files, and unsupported content.
- 02
Extract
Remove interface noise and rank source passages against the caller’s task.
- 03
Prove
Return source metadata, exact citations, content hashes, and injection warnings.
- 04
Reuse
Cache only eligible public content so repeated work increases margin.
THE PRODUCT IS THE RECEIPT
Enough evidence to trust—or reject—the result.
Every success exposes what was fetched, how it was transformed, what was omitted, estimated unit economics, and whether the page contained suspicious instructions. Settlement proof arrives in the x402 response header.
{
"schema_version": "context-receipt.v1",
"receipt_id": "rcpt_…",
"context_id": "ctx_…",
"context": {
"markdown": "# Source …",
"citations": [{ "id": "c1", "excerpt_sha256": "…" }],
"estimated_tokens_saved": 1840
},
"delivery": {
"status": "delivered",
"price_usd": 0.005,
"payment": { "settlement_receipt_header": "PAYMENT-RESPONSE" }
},
"safety": { "robots_policy": "allowed", "prompt_injection_signals": [] }
}