Access for AI tools
Quiet keeps Docusaurus's generated HTML, headings, canonical URLs, metadata and links. It does not add a client-only document shell.
Markdown and llms.txt
Use docusaurus-plugin-llms to generate a documentation index and readable Markdown files. Biel's docs use this plugin.
npm install --save-dev docusaurus-plugin-llms
plugins: [
[
"docusaurus-plugin-llms",
{
generateLLMsTxt: true,
generateLLMsFullTxt: true,
generateMarkdownFiles: true,
excludeImports: true,
removeDuplicateHeadings: true,
ignoreFiles: ["private/**", "archive/**"],
},
],
];
Review the generated files after npm run build. Confirm that MDX components contain useful text in their exports. Excluding a page from an index is not access control.
MCP discovery
Expose an MCP endpoint from your retrieval provider. Add a discoverable URL in the document head:
headTags: [
{
tagName: "link",
attributes: {
rel: "mcp",
href: "https://mcp.example.com/mcp",
},
},
];
Replace the example with a working endpoint. Keep credentials out of public URLs. A discovery link does not create an MCP server or connect a client automatically.
For Biel, copy the project endpoint from its configuration. Biel's MCP guide covers public and authenticated access.
Diagrams and reference
Keep a text explanation beside a diagram. Generate API reference from the same specification used by the product. These sources should remain useful without running an interactive widget.
Verify the output
- Read the page HTML with JavaScript disabled.
- Check canonical URLs, sitemap entries and heading order.
- Check export exclusions and links after every build.
- Keep examples, prerequisites and limits in the page text.
These checks improve access to your content. They do not guarantee rankings or citations by an AI tool.