---
name: video-download
description: Download a public video, clip, or audio track from TikTok, Instagram, Facebook, X, YouTube, Vimeo, Reddit, Twitch, SoundCloud, or Bilibili through the Videodownload MCP server. Use when the user asks to save, download, extract audio from, or inspect the formats of a public video URL.
---

# Video download

Use the Videodownload MCP server to turn a public video URL into a file. The
server does the extraction. Never install yt-dlp, scrape the page yourself, or
ask the user for platform cookies.

## When to use

- The user gives a public video, reel, short, clip, or track URL and asks to
  save it, download it, or extract the audio.
- The user wants to know the title, duration, thumbnail, or available formats
  before deciding.
- The user asks an agent to collect a public video for a project they own or
  have permission to use.

Do not use this skill for private, follower-only, paid, DRM protected, or live
content. Do not use it to bypass a login wall.

## Tools

- `resolve_video` reads metadata and formats. Free.
- `list_formats` returns the format list only. Free.
- `download_video` starts an async job. Costs one credit.
- `get_download` returns progress and the signed file URL.
- `cancel_download` stops a job and refunds the credit.
- `extract_audio` starts an MP3 job. Costs one credit.

## Workflow

1. Call `resolve_video` with the exact public URL. If the user has not chosen a
   format, show the title, duration, and two or three format options.
2. Call `download_video` with `format` set to `mp4`, `audio`, or `gif` and
   `quality` set to a value like `1080p`, `720p`, or `audio`.
3. Poll `get_download` every 2 seconds while the status is `queued`,
   `resolving`, `downloading`, or `processing`. Stop on `ready`, `failed`,
   `canceled`, or `expired`.
4. On `ready`, give the user the signed URL and state that it expires 15
   minutes after the job became ready.
5. On `failed`, report the error code and one concrete next step. Do not retry
   more than once unless the user asks.

## Error handling

- `PRIVATE_CONTENT` or `LOGIN_REQUIRED`: the post needs an account. Explain
  that the service does not accept cookies and stop.
- `DRM_PROTECTED`: refuse. Do not suggest workarounds.
- `UNSUPPORTED_PLATFORM`: read `supported platforms` from
  https://videodownloadmcp.com/api-reference and give the closest option.
- `QUOTA_EXCEEDED`: the free web quota is used. Suggest an API key.
- `INSUFFICIENT_CREDITS`: the account needs credits. Link to
  https://videodownloadmcp.com/pricing.
- `TIMEOUT`: retry once. If it fails again, report the platform and URL.
- `FILE_TOO_LARGE` or `DURATION_LIMIT_EXCEEDED`: explain the 500 MB and 30
  minute caps.

## Safety

- Only public content. Never ask for passwords, session cookies, or API keys in
  chat.
- Do not download content the user does not own or have permission to use.
- Do not strip or alter watermarks or attribution.
- Do not send the file to another service without the user asking.
- Do not print the API key in a response.

## Example

User: "Download this TikTok as MP4: https://www.tiktok.com/@user/video/123"

1. `resolve_video({"url": "https://www.tiktok.com/@user/video/123"})`
2. Show title and options, then
   `download_video({"url": "...", "format": "mp4", "quality": "1080p"})`
3. Poll `get_download({"job_id": "job_..."})`
4. Reply with the signed URL and the 15 minute expiry note.

User: "Just give me the audio."

1. `resolve_video` to confirm the post is public.
2. `extract_audio({"url": "..."})`
3. Poll and return the MP3 link.
