Nabchan MCP Server

🚀 nabchan-mcp-server

This is an MCP server that returns Nablarch-related information based on the information from Nablarch's official documentation.

⚠️ Important Note

This project is currently in the experimental stage and there is still much room for improvement.

🚀 Quick Start

You can easily try it out using Docker:

graph LR
  a[VSCode]
  b[nabchan-mcp-server<br>（Docker container）]
  a -->|Standard input-output communication| b

Add the following settings to VSCode:

{
  "mcp": {
    "inputs": [],
    "servers": {
      "nablarch-document": {
        "command": "docker",
        "args": [
          "run",
          "-i",
          "--rm",
          "--network",
          "none",
          "-e",
          "TRANSPORT=stdio",
          "ghcr.io/backpaper0/nabchan-mcp-server"
        ]
      }
    }
  }
}

Set GitHub Copilot Chat to proxy mode and ask questions related to Nablarch.

💻 Execution via ToolHive

When executing via ToolHive, run the following command:

thv run --name nabchan-mcp-server ghcr.io/backpaper0/nabchan-mcp-server -- --transport=stdio

🔧 Architecture

It adopts an architecture that runs in a local Python environment.

It builds an index using the Full-Text Search Extension of DuckDB and the morphological analysis library Lindera.

It extracts text from the HTML files of the Nablarch documentation, performs morphological analysis, and uses it as the target field for full-text search. In addition, information such as titles, summaries, and contents are provided in Markdown format and can be accessed through the API provided by the MCP server.

graph TD
   h[Nablarch's documentation<br>（HTML files）]
   c(Text)
   d(Summary)
   t(Title)
   m(Markdown)
   i[DuckDB]

   h -->|Extract text with BeautifulSoup| c
   c -->|Perform morphological analysis with Lindera| i
   c -->|Generate summary with LLM| d
   d --> i
   h -->|Extract title with BeautifulSoup|t
   t --> i
   h -->|Convert to Markdown format with html2text| m
   m --> i

The APIs provided by the MCP server are as follows:

• read_document
  • Returns the content of the specified Nablarch documentation URL converted to Markdown format.
• search_document
  • Searches the Nablarch documentation. The returned information includes:
    • Title
    • URL
    • Summary

🛠️ Development Information of nabchan-mcp-server

📋 Development Environment Requirements

• Python 3.11
• uv
• Git
• Docker

🔍 Index Building

uv run -m tools.build_index

💡 Usage Tip

If you haven't fetched the submodule content yet, run git submodule init and git submodule update first.

🧪 Search Test

uv run -m tools.search_document -q "Nablarch"

💻 Example VSCode Configuration for Development

Replace /path/to/nabchan-mcp-server with the actual path:

{
  "mcp": {
    "inputs": [],
    "servers": {
      "nablarch-document": {
        "command": "uv",
        "args": [
          "--directory",
          "/path/to/nabchan-mcp-server",
          "run",
          "-m",
          "nabchan_mcp_server.main"
        ]
      }
    }
  }
}

If using SSE transmission, the configuration is as follows:

{
  "mcp": {
    "inputs": [],
    "servers": {
      "nablarch-document": {
        "type": "sse",
        "url": "http://localhost:8000/sse"
      }
    }
  }
}

When using SSE, start the server by running the following command first:

uv run -m nabchan-mcp-server.main --transport sse --host localhost