Video Reader MCP - VS Code Extension

A VS Code extension that enables AI models (GitHub Copilot) to read and analyze video files using Progressive Context Enrichment principles.

Features

🎬 Video Analysis: Extract metadata, frames, and audio from video files
🚀 Auto-configuration: Automatically configures MCP server for GitHub Copilot Chat
📊 Progressive Context: Start with lightweight overviews, fetch details on demand
🎯 Token Efficient: Minimizes context consumption while maximizing utility

In GitHub Copilot Chat:

@workspace Analyze this video: /path/to/video.mp4

The AI will use progressive context enrichment:

Tool	Description	Token Cost
`get_video_overview`	Get video metadata and frame timestamps	~200 tokens
`get_video_metadata`	Technical specs: duration, resolution, fps	~100 tokens
`estimate_analysis_cost`	Estimate tokens before extracting frames	~50 tokens
`get_frame`	Extract a single frame at timestamp	~10K tokens
`get_frames_batch`	Extract multiple frames (max 5)	~50K tokens
`extract_audio`	Extract audio track	~50 tokens
`analyze_video_full`	Full video analysis (use sparingly)	~100K+ tokens

Setting	Default	Description
`videoReaderMcp.autoStart`	`true`	Auto-start MCP server on VS Code startup
`videoReaderMcp.logLevel`	`info`	Log level (error, warn, info, debug)
`videoReaderMcp.defaultFrameQuality`	`80`	JPEG quality for extracted frames (1-100)
`videoReaderMcp.maxFrameWidth`	`1920`	Maximum width for extracted frames

This extension implements the Progressive Context Enrichment pattern:

❌ Bad:  analyze_video_full → ALL frames → 100K+ tokens
✅ Good: get_video_overview → get_frame(t=30) → get_frame(t=90) → ~20K tokens

The AI starts with lightweight metadata and progressively fetches specific data as needed, keeping token usage minimal.

View detailed logs:

MIT - see LICENSE