NoteDiscovery

What gamosoft/NoteDiscovery MCP server does

gamosoft/NoteDiscovery MCP server exposes a NoteDiscovery knowledge base to MCP-compatible AI clients. The underlying application stores notes as Markdown files in folders, while its web interface provides features such as search, linked-note exploration, tagging, templates, and editing. The MCP integration gives an assistant access to the same note collection used by the application.

The supported workflows cover both retrieval and changes. An assistant can search for notes, read their contents, create new notes, edit existing notes, append ideas, apply tags, and organize notes into folders. The README also describes exploring relationships between notes through the knowledge graph. These capabilities make the server suitable for personal wikis, project documentation, meeting notes, journals, and other collections where the source files should remain under the operator’s control.

How it works

The MCP server is included with the NoteDiscovery project rather than being a separate hosted service. A Docker-based MCP configuration starts the NoteDiscovery image with python -m mcp_server. The container receives NOTEDISCOVERY_URL, which points to the running NoteDiscovery web application; the example uses http://host.docker.internal:8000.

This arrangement separates the note application from the MCP process while keeping them connected over the local deployment URL. MCP clients such as Claude Desktop and Cursor can start the process through their standard server configuration. The README identifies both clients as supported and also states that other MCP-compatible tools can use the integration.

Setup and configuration

First run NoteDiscovery, for example with the published Docker image mapped to port 8000 and a local data directory. The application’s notes are persisted in the mounted ./data/ directory according to the documented quick-start setup.

Then add the MCP server configuration to the client:

{
  "mcpServers": {
    "notediscovery": {
      "command": "docker",
      "args": [
        "run", "--rm", "-i",
        "-e", "NOTEDISCOVERY_URL=http://host.docker.internal:8000",
        "ghcr.io/gamosoft/notediscovery:latest",
        "python", "-m", "mcp_server"
      ]
    }
  }
}

The URL should identify the NoteDiscovery instance that contains the notes. The project also supports running the application directly with Python or through Docker Compose, but the documented MCP example uses the published container image. No paid API key is described for this integration.

Tools and capabilities

gamosoft/NoteDiscovery MCP server supports these documented capability groups:

• Search and discover notes using natural-language requests.
• Read note contents for retrieval and context.
• Create and edit Markdown notes.
• Append new ideas to existing notes, such as a daily journal.
• Tag and categorize notes, including finding notes with a tag.
• Organize notes into folders, including moving project notes to an archive.
• Explore connections represented in the knowledge graph.
• Use templates when creating structured notes.

The excerpt does not provide individual MCP tool names, schemas, argument lists, or return formats, so client-specific tool calls should be verified in the project’s MCP documentation.

Limitations and notes

The MCP integration depends on a running NoteDiscovery application and a reachable URL. It is designed for a self-hosted Markdown vault, not for a remote SaaS note provider. The README describes optional password protection for self-hosted deployments, but the supplied MCP example does not show an authentication setting or credential parameter.

The Docker configuration uses host.docker.internal, which is intended to let the MCP container reach an application running on the host. Network behavior can vary by operating system and Docker setup, so the endpoint may need to be adjusted for a particular deployment. The project is MIT-licensed and describes the client as using only the Python standard library, with no additional client dependencies.