products/docs-mcp

Duende Documentation MCP Server

MCP is an open protocol that enables AI models to securely interact with local and remote resources through standardized server implementations. This project contains an MCP server that can extend any large language model that supports MCP with up-to-date knowledge about Duende products, sourced from documentation, blog and samples.

Register the MCP

To run the Duende Documentation MCP Server, you will need the dnx tool (included in the .NET 10 SDK) in your system's PATH. The dnx tool can download and run applications packaged and distributed through NuGet.

You can then register the MCP server in your tool of choice. While the exact steps may vary depending on your IDE, here are some common configurations.

Visual Studio / Visual Studio Code

You can register the Duende Documentation MCP Server in your user settings to make it available in any workspace.

Alternatively, you can add a .vscode/mcp.json file to your workspace:

{
  "servers": {
    "duende-mcp": {
      "type": "stdio",
      "command": "dnx",
      "args": ["Duende.Documentation.Mcp", "--yes", "--", "--database", "/path/to/database.db"],
      "env": {}
    }
  }
}

The Duende Documentation MCP Server will create its database index at the path defined in the --database parameter.

Command-line Options

Flag	Description	Default
--database <path>	Fully qualified path to the SQLite database file	mcp.db (relative to working directory)
--with-http [port]	Enable the HTTP transport (Streamable HTTP) on the given port	Disabled; port defaults to 5800

Next, open GitHub Copilot and select Agent Mode to work with the MCP server.

JetBrains Rider

In JetBrains Rider settings, navigate to Tools | AI Assistant | Model Context Protocol (MCP). Next, add a new MCP server. In the dialog that opens, select As JSON and enter the following configuration:

{
  "mcpServers": {
    "duende-mcp": {
      "command": "dnx",
      "args": ["Duende.Documentation.Mcp", "--yes", "--", "--database", "/path/to/database.db"]
    }
  }
}

The Duende Documentation MCP Server will create its database index at the path defined in the --database parameter.

Claude Code

Execute the following command:

claude mcp add --transport stdio duende-mcp -- dnx Duende.Documentation.Mcp --yes -- --database /path/to/database.db

The Duende Documentation MCP Server will create its database index at the path defined in the --database parameter.

Tools and Example Prompts

The Duende Documentation MCP Server has several tools available:

• Free text search on blogs, docs, or samples
• Fetch specific page
• Get all content for a sample
• Get a specific file from a sample

The Duende Documentation MCP Server has instructions to announce the tools it provides, and instructs the LLM to use them. While this MCP prompt is elaborate, you may need to be explicit in prompts and for example, add "Use Duende samples" when you expect to update code with your LLM.

Example prompts:

• "What is a client in OpenID Connect?"
• "What is automatic key management?"
• "How can I validate a JWT token in ASP.NET Core?"
• "What is a Personal Access Token and how do I create one?"

Sometimes, it may be necessary to provide more context to the LLM. For example, when you want to know more about a specific topic you expect in the Duende documentation or blog, you can instruct the LLM to use the Duende Documentation MCP Server:

• "Explain .NET TLS certificates - use Duende"
• "Can I add passkeys to Razor Pages? use Duende"

Support

If you experience an issue with the Duende Documentation MCP or have any other feedback, please open an issue in our Duende community.

Disclaimer

Duende's AI developer tools (including the Duende Documentation MCP Server and Duende Agent Skills) are designed to provide Large Language Models (LLMs) with verified, structured context from Duende's documentation and product knowledge. These tools improve the quality and relevance of AI-assisted development with Duende products, including IdentityServer, BFF and our Open Source offerings, but they do not guarantee the correctness, security, or completeness of AI-generated output. All code, configuration, and architectural decisions produced with the assistance of these tools must be reviewed and validated by qualified developers before deployment to any environment. Duende Software is not responsible for AI-generated output that results from the use of these tools.

The Duende Documentation MCP Server provides LLM agents with access to Duende's official product documentation, website, and sample code. While this improves the accuracy of AI-assisted responses about Duende products, the LLM may still produce responses that are incomplete, outdated, or incorrect. The MCP Server serves context; it does not validate, execute, or enforce the correctness of the LLM's output. Always verify AI-generated guidance against the official documentation and samples at docs.duendesoftware.com.

Technical details

Development

• Run the project with --with-http 3000 to enable the HTTP transport (e.g., on port 3000).
• In VS Code, add a .vscode/mcp.json to your workspace:

  {
    "servers": {
      "duende-mcp": {
        "type": "http",
        "url": "http://localhost:3000"
      }
    }
  }
  

Indexers

The project uses full-text search with SQLite. There are indexes for docs, blog, and samples. Indexes are built by dedicated background services.

Docs

Documentation is indexed by parsing LLMs.txt hosted on Duende documentation: https://docs.duendesoftware.com/llms.txt

LLMs.txt is a technique that allows LLMs to find information in a Markdown-based format, so it can be parsed more easily (and within the LLM context). While available on many websites, none of the vendors currently support this out-of-the-box.

Blog

Blogs are indexed by parsing the RSS feed of the at https://duendesoftware.com/blog/, and then fetching each page and converting it into markdown.

Samples

Samples are indexed by looking at the samples-specific LLMs.txt file at https://docs.duendesoftware.com/_llms-txt/identityserver-sample-code.txt

This document contains sample names and descriptions, and includes links to GitHub. The GitHub repository is downloaded as an archive, and all .cs, .cshtml and relevant .js files are added to the index. Other files are ignored.