mcp s

MCP Servers provides a Docker-based image that exposes multiple MCP tools (such as a Git tool, a PostgreSQL-backed database tool, a DuckDuckGo-like search tool, and a knowledge-graph tool) plus an inspector UI on port 6274. After starting the container, you can access the inspector at http://localhost:6274/ to see which tools are available and how they are wired. The tools are exposed as SSE endpoints (e.g., http://localhost:9099/sse for the Git service, http://localhost:9098/sse for PostgreSQL, http://localhost:9097/sse for DuckDuckGo search, and http://localhost:9096/sse for knowledge graph). You can configure Cursor or other clients to connect to these endpoints by adding them to your MCP configuration as mcpServers entries. The container also maps your project directory to /project inside the container, enabling you to run commands against your local project data. The provided start script and guidance help you run the image and then interact with the tools via standard IO or the Cursor integration.

Prerequisites:

• Docker installed on your system
• A project directory you want to expose to MCP tools (this will be mapped to /project inside the container)
• Optional: a .env file and a .memory file in your project root

1. Pull and run the published image (recommended):

docker run -it --rm
-p 6274:6274 -p 6277:6277 -p 9099:9099 -p 9098:9098 -p 9097:9097 -p 9096:9096
-v ./:/project -v ./.memory:/memory
-e MEMORY_FILE_PATH=./.memory
--add-host=host.docker.internal:host-gateway
--env-file ./.env
--name mcp-servers ghcr.io/cmiles74/docker-mcp-servers:latest

This starts the container with the MCP servers exposed and ready to use. The project directory is mounted to /project inside the container and the inspector is available at port 6274.

2. Alternative: run locally from source (optional if you have access to the container scripts)

• Build or pull the image as per the repo's documentation
• Use the provided start script to boot the container and expose the necessary ports and volumes

3. Interacting with the tools

• Open http://localhost:6274/ to view the inspector and available MCP servers
• Add the MCP configuration to your Cursor setup or any client by using the provided JSON structure (see example in the repository)

Note: If you want to customize the database connection, ensure MCP_DATABASE_URI is provided in your environment (e.g., via an .env file) and that host.docker.internal is reachable from within Docker on your platform.

Tips and common issues:

• The Docker image exposes multiple MCP servers; not all may be fully tested. Use with caution in production.
• If you rely on a PostgreSQL MCP tool, ensure MCP_DATABASE_URI is set in the container environment and that the host is accessible as host.docker.internal from inside Docker.
• The memory store for the LLM data is configured via MEMORY_FILE_PATH; ensure the path exists and is writable.
• The inspector on port 6274 is useful for discovering which tools are available and how to wire them into Cursor or other clients.
• If you modify the .env file, remember to restart the container to apply changes.
• When integrating with Cursor, you can paste a JSON like the one in the README to expose the MCP servers to Cursor’s MCP menu.