fuma-nama/fumadocs fumadocs@16.15.9

on GitHub

• @fumadocs/base-ui@16.15.9
• fumadocs-core@16.15.9
• fumadocs-ui@16.15.9

getPageByUrl() on the loader

Look up a page by its URL:

source.getPageByUrl('/docs/getting-started');
source.getPageByUrl('/cn/docs/getting-started', 'cn');

Without the language argument every language is looked up, unlike getPageByHref() which resolves the default language only.

llms() renders pages

llms() used to build the llms.txt index only, turning a page into Markdown was left to your own getLLMText(). Pass renderPage and it covers both:

import { llms } from 'fumadocs-core/source';

export const docsLlms = llms(source, {
  renderPage: async (page) => `# ${page.data.title} (${page.url})

${await page.data.getText('processed')}`,
});

• page(page) renders one page, for the per-page Markdown route.
• full(lang?) renders every page and joins them, for llms-full.txt.

// app/llms-full.txt/route.ts
export const GET = async () => new Response(await docsLlms.full());

Both methods exist only when renderPage is given, in types and at runtime. Fumadocs cannot know how your content source exposes Markdown: page.data.getText('processed') on Fumadocs MDX, page.data.content on @fumadocs/local-md.

fumadocs-core/mcp: docs tools for your MCP server

Register the docs tools on a server you own, rather than on a handler Fumadocs builds for you:

import { createMcpHandler, McpServer } from '@modelcontextprotocol/server';
import { registerSearchTool, registerSourceTools } from 'fumadocs-core/mcp';
import { createFromSource } from 'fumadocs-core/search/server';
import { docsLlms, source } from '@/lib/source';

const handler = createMcpHandler(() => {
  const mcp = new McpServer({ name: 'docs', version: '1.0.0' });

  registerSourceTools(mcp, source, docsLlms);
  registerSearchTool(mcp, createFromSource(source));

  return mcp;
});

export const GET = (req: Request) => handler.fetch(req);
export const POST = (req: Request) => handler.fetch(req);
export const DELETE = (req: Request) => handler.fetch(req);

Each takes the integration it reads from, so you can add your own tools next to them, point the search tool at another server, or register only one of the two.

@modelcontextprotocol/server is an optional peer dependency, and registerSourceTools needs a llms() output with renderPage.

toDocuments() for Algolia and Orama Cloud

Build the search indexes of every page, instead of mapping pages by hand:

// app/static.json/route.ts
import { toDocuments } from 'fumadocs-core/search/algolia';
import { source } from '@/lib/source';

export const GET = async () => Response.json(await toDocuments(source));

It awaits structuredData when your collection is async (React Router and TanStack Start), a step that was easy to miss. Pass tag to filter results by a value of your choice:

toDocuments(source, { tag: (page) => page.slugs[0] });

Exported from fumadocs-core/search/algolia and fumadocs-core/search/orama-cloud.

Marked as side-effect free

fumadocs-core now declares "sideEffects": false. No module in the package imports for side effects or ships CSS, so bundlers that rely on the hint (webpack in particular) can drop unused modules instead of keeping them alive:

import { createGetUrl } from 'fumadocs-core/source';