zed -threadbridge

This MCP server is the ThreadBridge extension for Zed, providing persistent AI memory across conversations by using a hybrid semantic memory stack (Arctic Embed + BM25 + Reciprocal Rank Fusion) and a cognitive decay model (BLL). It exposes tools that let you load and save conversation context and perform semantic searches over project memory to retrieve relevant past interactions. In Zed, once the ThreadBridge extension is installed (either via the marketplace or local development flow), the MCP server can be launched using the standard container image. You can then invoke the MCP tools from within your Zed projects to manage context across sessions and projects. The available MCP tools are: load_thread (load saved conversation context for a project), save_thread (persist the current conversation context), and search_memory (semantically search across project memory). If you need to tailor which memory project is used, you can optionally configure a project_path in Zed settings; otherwise the server defaults to the current working directory.

Prerequisites:

• Docker installed and running
• Optional: Zed installed if you are integrating with the Zed extension workflow

Installation steps:

1. Decide how you want to run the MCP server. The recommended approach for this extension is to use the prebuilt Docker image.
2. Start the MCP server (ThreadBridge) using Docker: docker run -i dereklei12/mcp-threadbridge
3. If you prefer to build and run locally (advanced):
   • Ensure Rust is installed (recommended only for development of the extension itself, not required for end-to-end use).
   • Build the mcp-threadbridge binary from the repository, then run the binary directly as your MCP server.
4. In Zed, install the ThreadBridge extension via the Extensions Marketplace or follow the local development workflow described in the extension’s README. The extension will automatically fetch or build the appropriate binary on first use.
5. In your MCP client configuration, point to the running ThreadBridge server (for Docker this is typically via the command shown in mcp_config).

Tips and caveats:

• The ThreadBridge extension relies on persistent memory; ensure the container has access to storage or configure a volume if you require durable memory storage across restarts.
• If you update the extension, re-deploy the Docker container or rebuild the binary to pick up changes to the memory model or tools.
• The optional project_path setting allows you to scope memory to a specific project folder; if omitted, the server uses the current working directory.
• Common issues include Docker daemon not running, permission errors when binding volumes, or mismatched voltages/architectures on Windows. Check Docker Desktop logs and ensure the image is pulled successfully.
• Tools available from MCP exposure: load_thread, save_thread, and search_memory. Use these to manage and query memory across conversations.
• The server targets macOS (Apple Silicon), Linux x86_64, and Windows x86_64 platforms, but ensure your Docker image supports your host architecture.