Page Capture

POST /accounts/{account_id}/capture (accounts.write): a picture of one page, opened in the X account's own browser and proxy. Read-only: the browser opens the page in a new tab, hides cookie banners and X's own overlays with CSS (it never clicks them, so nothing is accepted or dismissed), takes a JPEG and closes the tab. Nothing is clicked or typed. It isn't a write: no Idempotency-Key, no receipt, not paced.

What It Opens#

Only these addresses, written exactly so. Anything else answers 400 errors/invalid_parameters (listing them) before any browser opens.

Hosts are matched exactly: x.com.evil.com, status.claude.com.evil.com or another subdomain is refused. status.anthropic.com and the bare githubstatus.com, vercel-status.com, cloudflarestatus.com and netlifystatus.com redirect elsewhere: use the address above. The browser checks the address again before it opens it and after the page loads, so a page that goes to another address is never captured (left_allowed_hosts).

X Posts#

The post is opened from the account's own signed-in X (never a signed-out view) and the picture is the post's own article on its page, found by X's markup and the post's own link, never by its words: no other post of the conversation, and the reply box and X's header, bars, sheets and floating buttons hidden. It ends at the post's action bar (reply, repost, like, bookmark, share), without X's sort row under it, and the author's Subscribe and Follow buttons are hidden, like a plain screenshot of the post. With x_post: { counts: false }, the post's time and views row and its action bar are hidden too, and the picture ends 16 px under its last content (its text, media or quoted post): no views, replies, reposts, likes or bookmarks. The post is captured only when X shows it under the handle in the address; otherwise nothing is captured:

Request#

curl -sS -X POST "https://api.lempile.com/accounts/janedemo/capture" \
  -H "X-API-KEY: $LEMPILE_API_KEY" \
  -H "content-type: application/json" \
  -d '{ "url": "https://x.com/OpenAI/status/1986000000000000001", "width": 520, "dpr": 2, "x_post": { "counts": false } }'

Answer#

The picture: 200, with the JPEG as base64 and its size in pixels (CSS size times dpr), the address the page ended on and its title.

{ "object": "PageCapture", "ok": true, "account_id": "11111111-1111-4111-8111-111111111111",
  "url": "https://x.com/OpenAI/status/1986000000000000001",
  "final_url": "https://x.com/OpenAI/status/1986000000000000001",
  "title": "OpenAI on X: \"…\" / X", "width": 1040, "height": 1224,
  "jpeg_base64": "/9j/4AAQSkZJRg…", "captured_at": "2026-11-07T18:20:00.000Z" }

No picture: still 200, with ok: false, the browser's code and detail. Nothing was clicked or typed; the code says whether trying again can help.

{ "object": "PageCapture", "ok": false, "account_id": "11111111-1111-4111-8111-111111111111",
  "url": "https://x.com/fakeopenai/status/1986000000000000001",
  "code": "x_post_author_mismatch",
  "detail": "X shows this post under @OpenAI, not @fakeopenai: nothing was captured.",
  "captured_at": "2026-11-07T18:20:00.000Z" }

Besides the X post codes above:

The account itself answers as on other routes: 423 errors/x_account_locked, 422 errors/proxy_auth_failed or errors/proxy_unreachable, 401 errors/disconnected_account. 400 errors/invalid_parameters is a field out of range or an address it doesn't open, 403 a key without accounts.write, 404 no such X account, 503 errors/service_unavailable the browser runner out of reach, and 504 errors/timeout no picture within 90 seconds. See Errors And Limits.