video-research

The video-research MCP server provides a suite of tools to analyze video content, perform web-enabled research, and organize results with a structured evidence framework. It exposes commands and agents that handle tasks like extracting frames from local videos, analyzing YouTube tutorials, conducting topic research with evidence grading, and grounding results to source documents. Use the server as a standalone MCP endpoint or integrate it with Claude or other MCP clients to perform workflows such as video analysis, web search, and document-backed research. Typical workflows include analyzing a video or URL to generate timestamped insights, creating concept maps, and storing results for recall and reuse across projects.

Prerequisites:

• Python 3.11 or newer
• uv (uvx) library
• Node.js v16+ only if you plan to use the npm-based installer wrapper
• A Google/Gemini API key (for Gemini-powered analysis)

Installation options:

Option A (recommended via npm wrapper):

1. Install and run the MCP server installer which configures 17 commands, 7 skills, and 7 agents to run via uvx.
2. Ensure GEMINI_API_KEY is available in your environment.

npx video-research-mcp@latest
export GEMINI_API_KEY="your-key-here"

Check status and install locally for a project-specific setup:

npx video-research-mcp@latest --check
npx video-research-mcp@latest --local

Uninstall if needed:

npx video-research-mcp@latest --uninstall

Option B (manual uvx-based run):

1. Install the Python package via PyPI and run via uvx as configured in the MCP manifest (the installer produces this configuration).
2. Ensure environment variables are set, notably GEMINI_API_KEY.

# Example manual run after installing the Python package
uvx video-research-mcp

Prerequisites reminder:

• uv installed and accessible in PATH
• Node.js 16+ if you rely on npm-based installers
• GEMINI_API_KEY available in environment

The installer copies the necessary 28 tools into your Claude project workspace (~/.claude/), and configures the MCP server to run via uvx from PyPI.

Environment variables: GEMINI_API_KEY is required for Gemini-powered features. You can also configure additional related keys (e.g., API endpoints or Weaviate if you enable semantic recall) as needed by your setup. The MCP configuration shown in this readme uses a placeholder for the GEMINI_API_KEY; replace it with your actual key or reference via your environment. Common issues include: not having uvx installed or not exporting GEMINI_API_KEY, and network access problems when the Gemini API is rate-limited or blocked. When running locally, ensure the working directory is writable and that the ~/.claude/ directory exists for command and agent deployment. If you plan to run multiple MCP servers, consider using distinct environment configurations and separate keys where appropriate.