# AfterImage

AfterImage is a shared copy of the public web for AI agents.
If a page was fetched recently, reuse that copy instead of scraping again.
Each copy has a timestamp and a sha256 hash of the raw bytes.

Base URL: https://afterimage.page
Do not scrape this file. Call the JSON endpoints below.

## 1. Get a key (human, once)

POST https://afterimage.page/v1/billing/checkout
Content-Type: application/json

{"pack":"starter"}

Packs: starter = $5 USD, builder = $20 USD.
The JSON returns api_key and checkout_url.
Open checkout_url, pay, then keep api_key secret.
Send it on every later request:

  Authorization: Bearer ak_live_…

HTTP 402 means the header is missing or the key has no credits left. Buy another pack.
GET https://afterimage.page/v1/billing/balance with the same header to see remaining dollars.

Prices:
- GET /v1/search = $0.005 per call
- GET /v1/page cache hit = $0.002 (copy still fresh)
- GET /v1/page live fetch = $0.01 (origin was contacted)

Optional: x402 USDC on Base instead of a key, if this instance advertises it
at GET https://afterimage.page/.well-known/x402

## 2. Search copies (does not visit the live web)

GET https://afterimage.page/v1/search?q=fastapi+background+tasks

Query:
- q (required): words to find in titles and text
- limit (optional, default 10, max 50)
- max_age_s (optional): only copies this many seconds old

Response:
- indexed: how many copies are stored
- hits[]: url, title, snippet, hash, fetched_at, age_s, status, score
If indexed is 0, the library is empty. Fetch URLs with /v1/page, then search again.
If hits is empty but indexed > 0, nothing matched. Try different words or fetch the URL.

## 3. Fetch or reuse a URL

GET https://afterimage.page/v1/page?url=https://example.com/pricing&max_age_s=900

Query:
- url (required): http or https. No logins, no private/internal addresses.
- max_age_s (optional, default 900): reuse a stored copy if it is no older than this

Response:
- cache: "hit" (reused) or "miss" (live fetch)
- text: readable extract, not HTML
- hash: sha256 of the raw response body
- fetched_at, age_s, status, title, final_url

If you already have the URL from a search hit, call /v1/page to get the full text.
If cache=hit, do not fetch the origin yourself unless you need a smaller max_age_s.

## MCP

POST https://afterimage.page/mcp
JSON-RPC. Tools: search_pages, get_page.
initialize and tools/list are free. tools/call for those two tools needs the same
Authorization header as HTTP, or x402 USDC. HTTP 402 if unpaid.

Server card: GET https://afterimage.page/.well-known/mcp.json
Also: GET https://afterimage.page/mcp/server-card

## Other

- Health: GET https://afterimage.page/health
- OpenAPI: GET https://afterimage.page/openapi.json
- Agent card: GET https://afterimage.page/.well-known/agent-card.json
- x402 discovery: GET https://afterimage.page/.well-known/x402