laradji/deadzone

What it does

Deadzone is a semantic search tool designed for searching through library documentation locally without requiring internet connectivity. It functions as an MCP (Model Context Protocol) server that provides vector-based semantic search capabilities over documentation, similar to Context7 but operating completely offline. The tool indexes documentation and enables intelligent natural language searching through prebuilt vector databases.

How to set up

Deadzone is distributed as a single Go binary, making installation straightforward. On first launch, it automatically fetches a prebuilt index database and verifies it using SHA256 checksums. After the initial setup, it operates with zero network requirements. Alternatively, you can run it via Docker using the command: docker run --rm -i ghcr.io/laradji/deadzone. The tool integrates with MCP stdio for seamless communication with AI assistants.

Key features

• Local-first architecture with zero network dependency after initial setup
• Semantic search using vector indices pinned to binary version
• Automatic prebuilt index database fetching and SHA256 verification
• Single Go binary deployment for easy distribution
• MCP server protocol support for AI assistant integration
• Privacy-preserving offline operation