HelpNDoc's Model Context Protocol (MCP) server allows compatible AI agents to connect to HelpNDoc and interact with the currently opened project. Using the tools exposed by the server, AI agents can autonomously inspect and manage project elements such as topics, keywords, tags, statuses, and library items.

The MCP server uses HTTP connections and can be used by any MCP-compatible client or AI agent that supports the HTTP transport.

Note: The MCP server can modify the currently opened HelpNDoc project. Carefully review the permissions, configuration, and actions of any AI agent before connecting it to HelpNDoc. It is also recommended to save or back up the project before allowing an autonomous agent to perform extensive changes.

Note: This feature might not be available on all editions of HelpNDoc. Check HelpNDoc's feature comparison page to learn more.

Configure the MCP server

The MCP server's listening port can be configured from HelpNDoc's options:

1. Click File, then click Options.
2. Select Artificial Intelligence (AI).
3. Select MCP Server.
4. Enter the port number in the Port field.
5. Click OK to save the changes.

The default port is 8933.

The configured port must be available before the MCP server can start. When changing the port, update the MCP connection settings in any AI agent that connects to HelpNDoc.

Start the MCP server

To start accepting connections from MCP-compatible AI agents:

1. Open the project that the AI agent will access.
2. Click the Tools ribbon tab.
3. Locate the Services group.
4. Click Start MCP server.

HelpNDoc starts listening for incoming HTTP connections on the configured port. The Start MCP server button changes to Stop MCP server, indicating that the server is running.

You can then configure the AI agent to connect to HelpNDoc using an HTTP MCP connection and the port configured in HelpNDoc's options.

Stop the MCP server

To stop accepting incoming connections:

1. Click the Tools ribbon tab.
2. Locate the Services group.
3. Click Stop MCP server.

HelpNDoc stops listening for incoming HTTP connections. The button changes back to Start MCP server.

Connected AI agents can no longer call HelpNDoc tools after the server has been stopped.

Using the MCP server

After connecting, an AI agent can discover and call the tools exposed by HelpNDoc's MCP server. These tools allow the agent to read project information and perform supported operations on the currently opened project.

Setup and usage instructions are available for the following agents:

• Using the HelpNDoc MCP server with Claude Code 
• Using the HelpNDoc MCP server with Codex application 
• Using the HelpNDoc MCP server with the Codex extension for Visual Studio Code 
• Using the HelpNDoc MCP server with GitHub Copilot in Visual Studio Code 

Depending on the capabilities and instructions of the connected agent, it can perform tasks such as:

• Inspecting the project's structure and metadata
• Reading or updating topic properties and content
• Creating, moving, or deleting topics
• Managing keywords and their topic associations
• Managing tags and statuses
• Inspecting and managing project library items
• Navigating to project elements in the HelpNDoc user interface

Operations are performed directly on the project currently opened in HelpNDoc. Save the project regularly and review changes made by autonomous agents.

For the complete list of available methods and details about their parameters and behavior, see: Tools available through the MCP server.

Connection requirements

The HelpNDoc MCP server only accepts HTTP connections. MCP clients configured exclusively for other transport methods cannot connect to it.

Before connecting an AI agent, verify that:

• The MCP server is running in HelpNDoc.
• The agent is configured to use HTTP.
• The configured port matches the port set in HelpNDoc.
• The port is available and not blocked by local security software.
• The correct HelpNDoc project is open.

Recommended precautions

AI agents can perform multiple operations autonomously and may modify significant parts of a project. Before allowing an agent to work on a project:

• Save the current project.
• Create a backup when performing extensive or experimental operations.
• Give the agent clear and limited instructions.
• Review the agent's proposed actions when the client supports approval prompts.
• Stop the MCP server when external access is no longer required.