CrabCut
Turn YouTube videos into short clips — from Claude, Cursor, or any AI assistant that supports MCP. You give it a YouTube link. It finds the best moments, reframes them for vertical video, adds subtitles, and gives you download links. All from a chat.
README
@crabcut/mcp-server
Turn YouTube videos into short clips — from Claude, Cursor, or any AI assistant that supports MCP.
Give it a YouTube link. It finds the best moments, reframes for vertical video (9:16), adds subtitles, and returns download-ready clips.
Quick Start
- Get an API key at app.crabcut.ai/developers
- Run:
CRABCUT_API_KEY=sk_live_... npx @crabcut/mcp-server
Setup
Claude Desktop / Cursor
Add to your config file — claude_desktop_config.json for Claude, .cursor/mcp.json for Cursor:
{
"mcpServers": {
"crabcut": {
"command": "npx",
"args": ["@crabcut/mcp-server"],
"env": {
"CRABCUT_API_KEY": "sk_live_..."
}
}
}
}
Cursor Directory
Smithery
npx @smithery/cli mcp add crabcut/crabcut-mcp-server
Remote (Streamable HTTP)
For n8n, custom integrations, or any client that supports remote MCP:
- Endpoint:
https://mcp.crabcut.ai/mcp - Auth:
Authorization: Bearer sk_live_... - Transport: Streamable HTTP (POST)
Tools
| Tool | What it does |
|---|---|
generate_clips |
Start clip generation from a YouTube URL. Returns a project_id right away. |
get_project_status |
Check progress and get clips when ready. Clips are sorted by score (best first). |
list_projects |
List your projects with status and clip counts. |
get_clip |
Get full details for a single clip. |
check_usage |
See your plan, remaining credits, and usage. |
How It Works
- Call
generate_clipswith a YouTube URL - Get back a
project_idimmediately - Poll
get_project_statusevery 10–15 seconds (or pass acallback_urlfor webhook) - When done, clips are sorted by score (best first), each with a
download_url
Statuses
Project: pending → processing → completed (or completed_no_clips / failed)
Processing steps: queued → DOWNLOADING_VIDEO → EXTRACTING_AUDIO → TRANSCRIBING → DETECTING_HIGHLIGHTS → CUTTING_CLIPS → EXPORTING_CLIPS → done
Clips: pending → exporting → completed (or failed). The download_url is available when clip_status is completed.
API Reference
generate_clips
Start clip generation from a YouTube video.
| Field | Type | Required | Description |
|---|---|---|---|
url |
string | Yes | Full YouTube video URL |
start_time |
number | No | Start time in seconds (to process only a segment) |
end_time |
number | No | End time in seconds (to process only a segment) |
callback_url |
string | No | Webhook URL to receive results when done |
{
"project_id": "43dbe622-8ac6-4579-9625-0ad7f0f9db0b",
"status": "processing",
"poll_url": "/api/v1/projects/43dbe622-8ac6-4579-9625-0ad7f0f9db0b",
"estimated_minutes": 5,
"message": "Processing started. Poll with get_project_status every 10-15 seconds until status is 'completed'."
}
get_project_status
Check progress and get clips.
| Field | Type | Required | Description |
|---|---|---|---|
project_id |
string | Yes | The project ID from generate_clips |
{
"id": "43dbe622-...",
"name": "Video Title",
"status": "completed",
"step": "done",
"error": null,
"expected_clips": 9,
"duration": 639,
"created_at": "2026-03-15T02:21:46.226Z",
"clips": [
{
"id": "32538b9c-...",
"title": "One Dating Theory Leads to Chaos",
"duration": 41.83,
"score": 90,
"reason": "Sharp universal joke that hooks instantly with strong reactions.",
"clip_status": "completed",
"download_url": "https://cdn.crabcut.ai/exports/premium/.../clip-32538b9c-....mp4",
"quality": "1080p",
"thumbnail_url": "https://cdn.crabcut.ai/exports/premium/.../clip-32538b9c-...-thumb.jpg",
"created_at": "2026-03-15T02:29:10.954Z",
"updated_at": "2026-03-15T02:30:43.907Z"
}
]
}
Clips are sorted by score (highest first).
list_projects
| Field | Type | Required | Description |
|---|---|---|---|
limit |
number | No | Max projects to return. Default 20, max 100. |
status |
string | No | Filter: pending, processing, completed, completed_no_clips, failed |
{
"projects": [
{
"id": "43dbe622-...",
"name": "Video Title",
"status": "completed",
"step": "done",
"expected_clips": 9,
"clips_count": 8,
"duration": 639,
"created_at": "2026-03-15T02:21:46.226Z"
}
],
"total": 1,
"limit": 20,
"offset": 0
}
get_clip
| Field | Type | Required | Description |
|---|---|---|---|
clip_id |
string | Yes | The clip ID from a project's clips array |
{
"id": "32538b9c-...",
"project_id": "43dbe622-...",
"title": "One Dating Theory Leads to Chaos",
"duration": 41.83,
"score": 90,
"reason": "Sharp universal joke that hooks instantly.",
"export_status": "completed",
"export_quality": "1080p",
"is_exported": true,
"video_url": "https://cdn.crabcut.ai/...",
"video_url_720p": "https://cdn.crabcut.ai/...",
"video_url_1080p": "https://cdn.crabcut.ai/...",
"thumbnail_url": "https://cdn.crabcut.ai/...",
"created_at": "2026-03-15T02:29:10.954Z",
"updated_at": "2026-03-15T02:30:43.907Z"
}
check_usage
No input required.
{
"plan": "pro",
"credits_remaining": 450,
"credits_total": 500,
"period_start": "2026-03-01T00:00:00.000Z",
"period_end": "2026-04-01T00:00:00.000Z"
}
Errors
| HTTP Status | Error | What to do |
|---|---|---|
| 400 | Invalid YouTube URL | Check the URL format |
| 401 | Unauthorized | Check your API key |
| 402 | Insufficient credits | Buy more credits or use a shorter video |
| 429 | Rate limit exceeded | Wait and try again |
| 500 | Internal server error | Try again later |
Webhooks
Pass a callback_url when calling generate_clips to receive a POST when all clips are ready.
Completed:
{
"event": "project.completed",
"project_id": "43dbe622-...",
"status": "completed",
"source_url": "https://www.youtube.com/watch?v=H51iLa1leOU",
"clips": [
{
"id": "32538b9c-...",
"title": "One Dating Theory Leads to Chaos",
"duration": 41.83,
"score": 90,
"download_url": "https://cdn.crabcut.ai/exports/premium/.../clip-32538b9c-....mp4",
"thumbnail_url": "https://cdn.crabcut.ai/exports/premium/.../clip-32538b9c-...-thumb.jpg",
"quality": "1080p"
}
]
}
Failed:
{
"event": "project.failed",
"project_id": "43dbe622-...",
"status": "failed",
"error": "YouTube video is unavailable",
"clips": []
}
Clips are sorted by score (highest first). We retry up to 3 times if your server doesn't respond.
Signature verification (optional): Each callback includes an X-Crabcut-Signature header. Compute HMAC-SHA256 of the request body using SHA-256(your_api_key) as the signing key. Compare with the header value (sha256=<hex>).
Using with n8n
Option A — Webhook (recommended):
- Create a workflow with a Webhook trigger node. Copy its URL.
- Add an HTTP Request node:
POST https://api.crabcut.ai/api/v1/clips/generatewith Header Auth (Authorization: Bearer sk_live_...) and body:{ "url": "https://www.youtube.com/watch?v=...", "callback_url": "https://your-n8n.com/webhook/abc123" } - When clips are ready, the Webhook node receives the payload.
- Best clip:
{{ $json.clips[0].download_url }}
Option B — Polling loop:
POSTto generate clips (same as above, withoutcallback_url).- Wait 15 seconds → GET
https://api.crabcut.ai/api/v1/projects/{{ $json.project_id }}→ IF status is notcompleted, loop back to Wait.
REST API
For direct HTTP calls without MCP:
| Method | Endpoint | Description |
|---|---|---|
POST |
/api/v1/clips/generate |
Start clip generation |
GET |
/api/v1/projects/:id |
Get project status and clips |
GET |
/api/v1/projects |
List projects |
GET |
/api/v1/clips/:id |
Get single clip details |
GET |
/api/v1/account/usage |
Check credits and plan |
Base URL: https://api.crabcut.ai — All endpoints require Authorization: Bearer sk_live_...
Environment Variables
| Variable | Required | Default |
|---|---|---|
CRABCUT_API_KEY |
Yes | — |
CRABCUT_API_URL |
No | https://api.crabcut.ai |
Links
- Website: crabcut.ai
- API key: app.crabcut.ai/developers
- Pricing: crabcut.ai/pricing
License
MIT
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。