Contendeo

What Contendeo MCP server does

The Contendeo MCP server turns supported video URLs into structured analysis for an MCP client. Its processing can include audio transcription, speaker identification, keyframe inspection, and OCR text extraction. The result is designed to preserve both what a video says and what appears on screen.

The service accepts URLs from platforms including YouTube, Instagram Reels, Vimeo, Twitter/X, TikTok, and direct video locations. It is useful for reviewing demonstrations, tutorials, dashboards, technical walkthroughs, design reviews, or other videos where visual information carries meaning that a transcript cannot capture.

How it works

Contendeo runs remotely at https://contendeo.app/mcp/, so the client sends tool requests to the hosted service rather than running the processing stack locally. The service downloads or fetches the video, extracts audio and keyframes, and applies the relevant analysis pipeline before returning a unified response.

quick_transcribe performs a faster, audio-only pass and returns a timestamped transcript with speaker identification. deep_analyze combines transcription, keyframe visual analysis, and OCR in one request. clip_context limits processing to a selected timestamp range, which is appropriate when only part of a long video matters. batch_analyze handles up to 10 videos in parallel and can provide synthesis across them.

Setup and configuration

The Contendeo MCP server requires no local package installation or downloaded model. Configure the remote MCP URL in a compatible client. Claude Desktop uses the following server entry:

{
  "mcpServers": {
    "contendeo": {
      "url": "https://contendeo.app/mcp/"
    }
  }
}

Claude Code can add the connection with claude mcp add --transport http contendeo https://contendeo.app/mcp/. Cursor, Continue, Cline, and other clients need support for remote MCP over streamable HTTP and should be pointed at the same endpoint.

Authentication uses OAuth 2.0. On first connection, the client redirects to Contendeo for Google OAuth or email sign-in; after approval, it receives a bearer token and reconnects. New accounts receive 10 free credits without requiring a card.

Tools and capabilities

• quick_transcribe: Produces a timestamped, speaker-aware transcript from audio.
• deep_analyze: Runs transcript generation, visual frame analysis, and OCR together.
• clip_context: Analyzes a specified time range rather than the entire video.
• batch_analyze: Processes up to 10 videos concurrently and supports cross-video synthesis.

Successful calls consume credits according to the selected operation. Quick transcription costs 1 credit, deep analysis costs 5, and clip analysis costs 1 for the quick mode or 3 for the deep mode. Batch processing is charged per video, with a 10% reduction at five or more videos. Cached results are free, and failed analyses are automatically refunded.

Limitations and notes

The Contendeo MCP server is a hosted service, not an open-source production backend. The MIT license applies to the wrapper repository’s documentation, examples, and manifest; the production server source is not open-sourced.

The free plan provides 10 signup credits. Paid options include monthly plans and pay-as-you-go credit purchases. The Power plan is the one listed with batch processing and webhooks. Actual processing also depends on the client supporting remote, streamable HTTP MCP connections and on the supplied video URL being fetchable by the service.