Use the docs with AI

Copy pages, open them directly in AI assistants, and connect the MCP server to keep your AI tools aligned with published documentation.

Every page in the Navixy docs includes built-in shortcuts for working with AI assistants. Use them to copy content, open pages directly in ChatGPT or Claude, or connect the docs directly to your AI tools through MCP.

Built-in page shortcuts

Shortcut	What it does	When to use it
Copy page	Copies the full page as Markdown	Paste full context into any AI assistant
View as Markdown	Opens the page as plain text in a new tab	Inspect the raw source or copy specific sections manually
Copy (on code blocks)	Copies a code block directly	Use for curl commands, JSON payloads, and config snippets to avoid selection errors
Open in ChatGPT	Opens the page in ChatGPT with content pre-loaded	Ask questions or summarize without pasting anything
Open in Claude	Opens the page in Claude with content pre-loaded	Ask questions or summarize without pasting anything
Export as PDF	Downloads the page as a PDF	Save a snapshot for offline use or share outside the docs

Using Open in ChatGPT and Open in Claude

Both shortcuts send the current page as context automatically. You can start prompting immediately without copying or pasting. This is useful when you want a quick answer about a specific page without switching tools manually.

Good prompts to start with:

• Summarize this page before I read it in full.

• Turn this into a checklist of prerequisites.

• Generate edge cases and test ideas from this reference.

• Explain what [parameter] does and when I'd change it.

Getting better output:

Add a constraint to shape the response. Ask for "steps only", "return JSON", or "assumptions first". If the page is long, specify which section you want the assistant to focus on.

Connecting the MCP server to your AI tools

The page shortcuts above work well when you already have the right page open. MCP (Model Context Protocol) goes further: it lets your AI tool search and retrieve any page from the Navixy docs on demand, without you needing to find and open the page first. This is especially useful during active development, when you're moving across multiple API references, device configurations, and integration guides in a single session.

Once connected, your AI tool retrieves answers directly from published documentation and can link to the specific sections it used.

MCP endpoint

The MCP server is public and requires no authentication:

https://www.navixy.com/docs/~gitbook/mcp

You can also copy this URL directly from the Connect with MCP option in the page shortcuts menu.

Connect MCP server

Use the configuration Below you can find configuration steps for most common tools. MCP works with any client that supports mcpServers configuration.

Claude Desktop

1. In Claude, go to Settings and select Developer.

2. Select Edit Config to open claude_desktop_config.json, then add the following:

{
  "mcpServers": {
    "navixy-docs": {
      "command": "npx",
      "args": [
        "mcp-remote",
        "https://www.navixy.com/docs/~gitbook/mcp"
      ]
    }
  }
}

1. Quit Claude completely and start it again. A window reload is not enough.

Cursor

VS Code

Other IDEs

Validate the MCP connection

After setup, run this prompt to confirm the server is responding:

Use the docs MCP server navixy-docs.
What pages cover IoT Logic flows?
List the pages you found and link each one.

If the response includes links to specific Navixy doc pages, the connection is working. Once confirmed, you can ask the assistant to fetch any page, find endpoints, or trace a full workflow across multiple doc sections.

Troubleshooting

Nothing changed after adding the server. Restart the tool completely. Quit the application and open it again. A window reload is not enough.

The assistant answers but does not cite sources. Ask for links explicitly. You can also ask it to quote a short phrase from the section it referenced to confirm it retrieved the right content.

Pasted content looks broken. Check for wrapped lines. This is most common with curl commands and long API paths copied from the docs.