# ansi — download videos by link

> API for AI agents: YouTube, TikTok (incl. links from modded apps), Instagram, VK, Twitch, Pinterest, Rutube and 1800+ sites.
> Send `X-Lang: ru` to get error messages in Russian (English by default).

## Get a key (two requests, email access required)
1. POST https://ansi.website/api/v1/auth/register  {"email": "...", "password": "at least 8 characters"}
   -> an email with a 6-digit code is sent (valid for 15 min)
2. POST https://ansi.website/api/v1/auth/verify    {"email": "...", "code": "123456"}
   -> {"api_key": "ansi_...", "session_token": "sess_...", "daily_limit": N}
No code? POST https://ansi.website/api/v1/auth/resend {"email"}. Disposable mailboxes are rejected.
Existing account: POST https://ansi.website/api/v1/auth/login {"email", "password"} -> session_token,
then POST https://ansi.website/api/v1/account/keys {"name": "..."} with Authorization: Bearer <session_token>
Pass the key as: Authorization: Bearer <key>  (or X-API-Key, or ?api_key=)
Without a key the REST API works anonymously: 40 requests/hour per IP. With an account: 300/day for all its keys.
Requests that fail through no fault of yours (no link found, site temporarily refusing, no subtitles) are not counted.
Keys can be created/revoked only with the session_token, not with an API key.

## Errors
Always JSON {"error": "human-readable message"}. 401 — no/invalid key, 403 — not allowed, 404 — not found/expired,
422 — bad input or the video can't be processed, 429 — limit reached (see Retry-After), 503 — site temporarily refusing, retry later.

## MCP (recommended for agents)
URL: https://ansi.website/mcp  (Streamable HTTP), header Authorization: Bearer <key>
Tools: get_video_info(url), download_video(url, quality, format), get_direct_links(url, quality, format),
       get_transcript(url, lang, format)

## Terminal
curl -s https://ansi.website/me -H "Authorization: Bearer <key>"   -> colored profile: usage bar, 7-day chart, keys (?plain=1 — no colors)

## REST
POST https://ansi.website/api/v1/info      {"url": "..."}   -> also "details": description, tags, views, likes, chapters, subtitle languages
POST https://ansi.website/api/v1/info/batch {"urls": ["...", "..."]}   -> up to 10 links, each counts as one request
POST https://ansi.website/api/v1/transcript {"url": "...", "lang": "en", "format": "text|segments|srt|vtt"}
  -> subtitles or auto-captions WITHOUT downloading the video (great for summarizing)
GET  https://ansi.website/api/v1/transcript?url=...&format=srt&api_key=...   -> plain text body
POST https://ansi.website/api/v1/jobs      {"url": "...", "quality": "1080", "format": "mp4", "webhook_url": "https://..."}
  -> {"job_id", "status": "queued"}; GET https://ansi.website/api/v1/jobs/<job_id> until status = done | error;
     webhook_url (optional, public http/https address) receives the same JSON via POST when the job finishes
POST https://ansi.website/api/v1/download  {"url": "...", "quality": "720", "format": "mp4", "mode": "link"}
  quality: best|2160|1440|1080|720|480|360|240   format: mp4|webm|mp3|m4a
  mode: link -> {"download_url": "...", "filename": "...", "size_bytes": N, "expires_in_sec": N}
        file -> the file itself; direct -> direct CDN links, no proxying
GET  https://ansi.website/api/v1/download?url=...&quality=720&format=mp4&api_key=...  -> file (curl -OJ)
GET  https://ansi.website/api/v1/account -> limit, today's usage, keys

OpenAPI docs: https://ansi.website/docs , https://ansi.website/openapi.json