gobbler

Gobbler is a universal content converter that turns diverse content types—YouTube transcripts, web pages, documents, audio/video, and even browser sessions—into clean Markdown with YAML frontmatter. The MCP interface exposes Gobbler as a set of programmable commands that AI agents and assistants can invoke. As a human user, you can operate Gobbler via the CLI (gobbler ... commands) or integrate it into automated workflows via the MCP server. The CLI supports converting videos, audio, documents, and web pages, and can batch process content like YouTube playlists or directories of files. The output is a consistent Markdown format that includes metadata such as source, type, title, duration, word_count, and a timestamp, making it readily consumable by AI agents and tooling that expect structured content. The MCP setup allows Claude Desktop/OpenCode-style agents to trigger Gobbler’s functionality through an mcp server configuration, enabling automated content extraction within AI workflows. Tools available include: gobbler youtube for transcripts, gobbler video/audio for speech-to-text, gobbler document for PDF/DODX/PPTX/XLSX, gobbler webpage for JS-rendered pages, and gobbler batch for batch processing; plus browser automation and a suite of Skills to guide agent usage. The MCP entry points align with the examples shown in the README, enabling agents to start Gobbler’s MCP server and issue commands like mcp add gobbler-mcp with uv to run the Gobbler MCP handler.

Prerequisites:

• Python 3.11+
• uv (Python package manager) installed via pipx/uv
• Docker Desktop (for web/document conversions via the backend backends)
• ffmpeg (for audio extraction from video)

Installation steps:

1. Clone the Gobbler repository: git clone https://github.com/Enablement-Engineering/gobbler.git

2. Install Gobbler locally (per project’s guidance): cd gobbler make install

3. If you plan to run the Docker-backed services for web/document conversions, start them: make start-docker

4. Run Gobbler MCP integration (example for Claude/OpenCode workflow):

   • Ensure uv is installed in your Python environment
   • Run the MCP command to register Gobbler as an MCP server (as shown in the examples): uv --directory /path/to/gobbler run gobbler-mcp

5. Verify installation by invoking a simple conversion through the CLI to confirm the pipeline is functional, for example: gobbler youtube "https://youtube.com/watch?v=..." -o output.md

6. Optional: follow the full setup from the Quick Start in the README to install and start all services, and to begin converting multiple content types.

Note: Replace /path/to/gobbler with the actual path to your gobbler project folder, and adjust commands as needed for your environment.

Tips and considerations:

• The output uses YAML frontmatter along with Markdown content, which is friendly for AI reasoning and downstream processing.
• For MCP integration, Gobbler exposes a local server that can be invoked by agents via uv; ensure the directory path is correct and that the gobbler-mcp entry point is present in your installation.
• When using Docker backends (Docling/Crawl4AI), ensureDocker Desktop is running and that any required Docker images are pulled/permitted by your environment.
• FFmpeg is required for audio extraction; verify ffmpeg is on your system PATH or accessible to Gobbler.
• If you encounter issues with content behind logins or JS-rendered pages, use gobbler webpage with appropriate URL flags, and ensure the browser automation permissions are configured via the Gobbler browser extension as described in the README.
• The MCP section examples show command structures for Claude Code/OpenCode; adapt the mcpServers configuration to match your agent’s MCP expectations and directory layout.
• When using batch features, ensure your input patterns (e.g., directory/file glob) are correctly specified and that output destinations exist or are writable.
• If you plan to automate via scripts, consider pinning Gobbler to a stable version and documenting the exact CLI invocations in your automation scripts.