MCP Video Parser

MCP server that details video clips for query

python

3

GitHub Stars

python

Language

6 months ago

First Indexed

2 months ago

Catalog Refreshed

Documentation & install

Readme and setup notes from the catalogue, plus a client-ready config you can copy for your MCP host.

Installation

Add the following to your MCP client configuration file.

Configuration

View docs

{
  "mcpServers": {
    "michaelbaker-dev-mcpvideoparser": {
      "command": "python",
      "args": [
        "mcp_video_server.py",
        "--http",
        "--host",
        "localhost",
        "--port",
        "8000"
      ]
    }
  }
}

You use this MCP server to analyze video content with AI, search transcripts and scenes, and query video data through natural language and time-based filters. It integrates vision models and chat capabilities to let you explore video collections efficiently while maintaining context across queries.

How to use

Start by running the MCP server locally and connecting with an MCP client. You can process videos, ask questions about specific moments, and perform location- and time-based searches. Use natural language queries to find relevant clips, request summaries, or drill into transcripts and scene changes. You’ll interact with an HTTP-style MCP endpoint from your client, and you can chat with integrated LLMs to refine results while keeping video context.

How to install

# prerequisites
python3.10+
# ensure Ollama is installed and running
# ensure ffmpeg is available for video processing

# 1. Clone the project
git clone https://github.com/michaelbaker-dev/mcpVideoParser.git
cd mcpVideoParser

# 2. Install dependencies
pip install -r requirements.txt

# 3. Pull required Ollama models (vision and chat)
ollama pull llava:latest
ollama pull mistral:latest

# 4. Start the MCP server (http endpoint on port 8000)
python mcp_video_server.py --http --host localhost --port 8000

Configuration and usage notes

Customize how you extract and analyze video data by adjusting configuration options. You can set the frame extraction rate, scene detection sensitivity, storage locations, and the LLM models used for vision and chat. The configuration is typically found in a default JSON file and can be edited to fit your environment and performance requirements.

Troubleshooting

Common issues and solutions help you get back on track quickly. If Ollama isn’t running, start it with ollama serve. If models are missing, pull them with ollama pull llava:latest and ollama pull mistral:latest. If the port is already in use, run the server on a different port, for example python mcp_video_server.py --http --port 8001.

Security and deployment notes

When deploying, consider restricting access to the MCP HTTP endpoint to trusted networks, enable authentication if available, and ensure you have proper storage permissions for video data and transcripts. Regularly update models and dependencies to keep the analytics accurate and secure.

Core MCP tools and capabilities

The server exposes a suite of MCP tools to analyze and query videos. These tools include processing videos, querying by location and time, searching content and transcripts, generating summaries, interacting with video-specific questions, analyzing moments, retrieving system statistics, and obtaining usage guidance.

Available tools

process_video

Process and analyze a video file to extract frames, transcripts, and metadata using vision models.

query_location_time

Query videos by location and time to retrieve relevant clips or summaries.

search_videos

Search video content and transcripts using natural language queries.

get_video_summary

Generate an AI-derived summary for a specific video.

ask_video

Ask questions about a particular video and receive context-aware answers.

analyze_moment

Analyze a specific timestamp within a video to extract details or identify events.

get_video_stats

Return system statistics and usage metrics for the MCP server.

get_video_guide

Provide usage instructions and guidance for operating the MCP server.