Deadzone
Give your agent offline semantic search across library documentation without sending queries or keys to a cloud doc API.
Overview
Deadzone is an MCP server for the Build phase that runs local semantic search over library documentation with no cloud API key or telemetry.
What is this MCP server?
- Local-first MCP: semantic search over bundled library docs
- No cloud, no API key, no telemetry per publisher description
- OCI image ghcr.io/laradji/deadzone:v0.7.4 (~230 MB) with baked deadzone.db and ONNX embedding assets
- Stdio transport; positional arg server for container launch
- Linux amd64/arm64 distroless base (publisher build_info v0.7.4)
- Version 0.7.4
- Approximate image size 230 MB
- Baked assets: deadzone.db, onnx_runtime, embedding_model
Community signal: 2 GitHub stars.
What problem does it solve?
Cloud doc tools and generic web search leak context, need keys, and break when you are offline or cannot send proprietary stack details upstream.
Who is it for?
Privacy-conscious solo developers who want bundled library docs and semantic search inside MCP without external retrieval SaaS.
Skip if: Builders who need up-to-the-minute docs for every npm package via live internet crawl, or who only develop on Windows without a Linux container runtime.
What do I get? / Deliverables
After you run the Deadzone OCI image with stdio MCP, your agent retrieves relevant library doc passages locally while you implement features.
- On-device semantic matches from bundled library documentation
- MCP-accessible doc snippets without cloud credentials
- Repeatable stdio integration using baked deadzone.db assets
Recommended MCP Servers
Journey fit
Reference lookup while implementing libraries is core Build work, especially when you want privacy and no extra SaaS billing. Docs is the shelf because Deadzone indexes library documentation for semantic retrieval, not deploy hooks or runtime monitoring.
How it compares
Offline baked-doc semantic search MCP, not a hosted documentation browser or general web search plugin.
Common Questions / FAQ
Who is Deadzone for?
Solo builders and small teams using MCP agents who want library documentation retrieval on-machine without API keys, cloud calls, or telemetry.
When should I use Deadzone?
Use it during Build when you are integrating third-party libraries and need fast, semantic doc lookup in Claude Code or Cursor without exposing your repo to external doc services.
How do I add Deadzone to my agent?
Pull ghcr.io/laradji/deadzone:v0.7.4, run the container with stdio MCP and the server positional argument, and register that process in your MCP client config.