Veedcrawl Mcp

Veedcrawl MCP Server

Give your AI agent eyes on the entire social video web — not just individual links.

The official Model Context Protocol server for Veedcrawl. Drop it into Claude, Cursor, or any MCP-compatible host and your agent can search social videos, pull metadata and transcripts, run structured extraction, and inspect public creator profiles across YouTube, TikTok, Instagram, X/Twitter, and Facebook.

---

What your agent can do

| Tool | What it returns | |---|---| | search_social_videos | Topic, keyword, creator, or trend search results with normalized video metadata | | get_video_metadata | Title, description, author, view/like/comment counts, tags, duration, thumbnails | | get_video_transcript | Full spoken transcript with timestamps — uses native captions or AI-generated speech-to-text | | extract_from_video | Structured answer to any prompt: hooks, CTAs, sentiment, topics, claims, quotes — guided by optional JSON Schema | | get_tiktok_profile | Creator bio, total video count, and up to 50 recent TikToks with full engagement stats per video | | get_instagram_profile | Creator bio, verified status, follower/following/post counts, and up to 50 recent posts (video, image, carousel) with per-post stats |

Why discovery tools change everything

Every other video API requires a public video URL. That leaves an agent stuck doing retrieval for content already found.

search_social_videos, get_tiktok_profile, and get_instagram_profile unlock discovery. Search a category or inspect a creator, select the videos that matter, then chain them into transcripts and extraction without collecting URLs by hand.

Find videos by topic, then analyze the winners:

search_social_videos(query="skincare morning routine", sort="most_liked", timePeriod="past_month")
→ normalized TikTok, Instagram, and YouTube video results
→ get_video_transcript(url=top_video.url)
→ extract_from_video(url=top_video.url, prompt="Extract the hook, claims, and CTA")

Audit a creator from scratch:

get_instagram_profile(username="hubermanlab", limit=50)
→ 283M followers, verified, recent posts with view/like counts
→ get_video_transcript(url=top_video.url)
→ extract_from_video(url=top_video.url, prompt="List every health claim made and the evidence cited")

Competitive analysis with no prep work:

get_tiktok_profile(username=competitor_A, limit=50)
get_tiktok_profile(username=competitor_B, limit=50)
get_tiktok_profile(username=competitor_C, limit=50)
→ Compare posting frequency, avg views, top hashtags, engagement rates across all three

---

Installation

npm install -g @veedcrawl/mcp

Get your API key at veedcrawl.com.

---

Setup

Claude Desktop

Edit ~/Library/Application Support/Claude/claude_desktop_config.json (macOS) or %APPDATA%\Claude\claude_desktop_config.json (Windows):

{
  "mcpServers": {
    "veedcrawl": {
      "command": "veedcrawl-mcp",
      "env": {
        "VEEDCRAWL_API_KEY": "ma_your_key_here"
      }
    }
  }
}

Restart Claude Desktop. The six tools will appear automatically.

Claude Code

claude mcp add veedcrawl -- env VEEDCRAWL_API_KEY=ma_your_key_here veedcrawl-mcp

Cursor

Add to your Cursor MCP config (~/.cursor/mcp.json):

{
  "mcpServers": {
    "veedcrawl": {
      "command": "veedcrawl-mcp",
      "env": {
        "VEEDCRAWL_API_KEY": "ma_your_key_here"
      }
    }
  }
}

Windsurf / Codeium

{
  "mcpServers": {
    "veedcrawl": {
      "command": "veedcrawl-mcp",
      "env": {
        "VEEDCRAWL_API_KEY": "ma_your_key_here"
      }
    }
  }
}

---

Tools reference

search_social_videos

Search public videos by topic, keyword, creator, or trend. Polling and retries are handled internally so the tool returns the completed search job with normalized video metadata.

| Parameter | Type | Required | Description | |---|---|---|---| | query | string | ✓ | Search phrase, topic, keyword, creator name, or trend | | platforms | string[] | — | Any of tiktok, instagram, youtube, facebook; defaults to TikTok, Instagram, and YouTube | | limit | number | — | Results to return: 1–100, default 25 | | sort | relevant \| latest \| most_liked | — | Result ordering | | timePeriod | string | — | all, past_24h, past_week, past_month, last_3_months, or last_6_months | | filters | object | — | Optional country, language, duration, likes, platform feature, or Facebook location filters |

Each result includes a public video URL you can pass directly into get_video_metadata, get_video_transcript, or extract_from_video.

---

get_video_metadata

Fetch structured metadata for any public video URL.

| Parameter | Type | Required | Description | |---|---|---|---| | url | string | ✓ | Public video URL (YouTube, TikTok, Instagram, X/Twitter, Facebook) |

Returns: platform, title, description, author info, view/like/comment/share counts, duration, thumbnails, tags, publish timestamp.

---

get_video_transcript

Return the complete transcript for a video. Polling and retries are handled internally — your agent gets the finished result.

| Parameter | Type | Required | Description | |---|---|---|---| | url | string | ✓ | Public video URL or direct media file URL | | mode | auto \| native \| generate | — | native = platform captions only · generate = AI speech-to-text only · auto = captions with AI fallback (default) | | lang | string | — | Language hint, e.g. en, es, ur |

---

extract_from_video

Ask any question about a video and get a structured answer. Optionally enforce output shape with a JSON Schema.

| Parameter | Type | Required | Description | |---|---|---|---| | url | string | ✓ | Public video URL or direct media file URL | | prompt | string | ✓ | What to extract — "List the key claims", "What is the hook?", "Summarize in three sentences" | | lang | string | — | Language hint for transcription | | schema | object | — | JSON Schema to constrain the output structure |

Example prompts that work well:

• "Extract the hook, main argument, and call to action"
• "List every product mentioned with the timestamp it appears"
• "What claims does the speaker make? Rate each one as factual, opinion, or unverified"
• "Rewrite this as a Twitter thread"

---

get_tiktok_profile

Fetch a TikTok creator's public profile snapshot. Returns author info, total video count, and their most recent videos with full engagement stats per video.

| Parameter | Type | Required | Description | |---|---|---|---| | username | string | ✓ (or url) | TikTok username, with or without @ | | url | string | ✓ (or username) | Full TikTok profile URL | | limit | number | — | Videos to return: 1–50, default 12 |

Each video in the response is a full object with views, likes, comments, shares, caption, hashtags, thumbnail, and publish timestamp — ready to pass directly into get_video_transcript or extract_from_video.

---

get_instagram_profile

Fetch an Instagram creator's public profile snapshot. Returns verified status, real follower and following counts, total post count, and recent content.

| Parameter | Type | Required | Description | |---|---|---|---| | username | string | ✓ (or url) | Instagram username, with or without @ | | url | string | ✓ (or username) | Full Instagram profile URL | | limit | number | — | Posts to return: 1–50, default 12 |

Post type is one of video, post (image), or sidecar (carousel). Video posts include a direct video URL you can pass into get_video_transcript.

---

Platforms supported

YouTube · TikTok · Instagram · X / Twitter · Facebook

---

Environment variables

| Variable | Required | Default | Description | |---|---|---|---| | VEEDCRAWL_API_KEY | ✓ | — | Your Veedcrawl API key. X_API_KEY is also accepted. | | VEEDCRAWL_BASE_URL | — | https://api.veedcrawl.com | Override for self-hosted or staging | | VEEDCRAWL_POLL_INTERVAL_MS | — | 1500 | Milliseconds between job status polls | | VEEDCRAWL_MAX_POLL_ATTEMPTS | — | 120 | Max poll attempts before timeout (~3 min) |

---

Links

• Get an API key
• Full API docs
• npm package

Source & license

This open-source MCP server is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: Veedcrawl
• Source: Veedcrawl/mcp
• License: MIT
• Homepage: https://veedcrawl.com

Install and usage instructions live in the source repository linked above.