AI integration

fromsrc provides built-in endpoints that make your documentation accessible to AI tools and LLMs.

llms.txt

The /llms.txt endpoint returns all your documentation content in a single file optimized for AI consumption.

https://your-docs.com/llms.txt

This endpoint combines all pages into one text file with clear section markers.

Per-page endpoints

Access raw markdown for any page by appending .md to the URL:

/docs/installation     → rendered page
/docs/installation.md  → raw markdown
/docs                 → rendered index page
/docs.md              → raw index markdown

This lets AI tools fetch specific pages without parsing HTML.

Use cases

RAG systems

Feed documentation into retrieval-augmented generation pipelines

AI assistants

Let chatbots answer questions using your docs

Code generation

Provide context for AI coding tools

Training data

Use docs as training material for fine-tuning

MCP endpoint

The /api/mcp endpoint supports both plain method calls and JSON-RPC 2.0 payloads.

json

{
  "jsonrpc": "2.0",
  "id": "1",
  "method": "search_docs",
  "params": { "query": "installation" }
}

Available methods:

initialize
notifications/initialized
ping
search_docs
get_page
list_pages
tools/list
tools/call
resources/list
resources/templates/list
resources/read

tools/call example:

json

{
  "jsonrpc": "2.0",
  "id": "2",
  "method": "tools/call",
  "params": {
    "name": "get_page",
    "arguments": { "slug": "installation" }
  }
}

tools/list and list_pages support optional pagination params:

json

{
  "jsonrpc": "2.0",
  "id": "3",
  "method": "tools/list",
  "params": { "cursor": "search_docs", "limit": 20 }
}

resources/read example:

json

{
  "jsonrpc": "2.0",
  "id": "4",
  "method": "resources/read",
  "params": { "uri": "fromsrc://docs/installation" }
}

Resource slugs are normalized lowercase kebab-case paths (for example fromsrc://docs/getting-started).

Implementation

Route handler

Create app/llms.txt/route.ts:

typescript

import { getAllDocs, getDoc } from "fromsrc";

export async function GET() {
  const docs = await getAllDocs("docs");
  const sections: string[] = [];

  for (const meta of docs) {
    const slug = meta.slug ? meta.slug.split("/") : [];
    const doc = await getDoc("docs", slug);
    if (doc) {
      sections.push(`# ${doc.title}\n\n${doc.content}`);
    }
  }

  return new Response(sections.join("\n\n---\n\n"), {
    headers: { "Content-Type": "text/plain" },
  });
}

Raw markdown endpoint

Use Next.js proxy to serve .md files:

typescript

// proxy.ts
import { NextRequest, NextResponse } from "next/server";

export function proxy(request: NextRequest) {
  const { pathname } = request.nextUrl;
  const accept = request.headers.get("accept") ?? "";
  if (pathname === "/docs.md" || pathname.endsWith(".md")) {
    const slug = pathname.replace("/docs/", "").replace(".md", "");
    return NextResponse.rewrite(new URL(`/api/raw/${slug}`, request.url));
  }
  if (pathname.startsWith("/docs/") && accept.includes("text/markdown")) {
    const slug = pathname.replace("/docs/", "");
    return NextResponse.rewrite(new URL(`/api/raw/${slug}`, request.url));
  }
  return NextResponse.next();
}

Best practices

Structure content clearly

Use descriptive headings and short paragraphs. AI tools parse structure to understand context.

Include code examples

Practical examples help AI understand usage patterns and generate accurate code.

Keep pages focused

One topic per page makes retrieval more precise and reduces noise in AI responses.

fromsrc provides built-in endpoints that make your documentation accessible to AI tools and LLMs.

llms.txt

The /llms.txt endpoint returns all your documentation content in a single file optimized for AI consumption.

https://your-docs.com/llms.txt

This endpoint combines all pages into one text file with clear section markers.

Per-page endpoints

Access raw markdown for any page by appending .md to the URL:

/docs/installation     → rendered page
/docs/installation.md  → raw markdown
/docs                 → rendered index page
/docs.md              → raw index markdown

This lets AI tools fetch specific pages without parsing HTML.

Use cases

RAG systems

Feed documentation into retrieval-augmented generation pipelines

AI assistants

Let chatbots answer questions using your docs

Code generation

Provide context for AI coding tools

Training data

Use docs as training material for fine-tuning

MCP endpoint

The /api/mcp endpoint supports both plain method calls and JSON-RPC 2.0 payloads.

json

{
  "jsonrpc": "2.0",
  "id": "1",
  "method": "search_docs",
  "params": { "query": "installation" }
}

Available methods:

initialize
notifications/initialized
ping
search_docs
get_page
list_pages
tools/list
tools/call
resources/list
resources/templates/list
resources/read

tools/call example:

json

{
  "jsonrpc": "2.0",
  "id": "2",
  "method": "tools/call",
  "params": {
    "name": "get_page",
    "arguments": { "slug": "installation" }
  }
}

tools/list and list_pages support optional pagination params:

json

{
  "jsonrpc": "2.0",
  "id": "3",
  "method": "tools/list",
  "params": { "cursor": "search_docs", "limit": 20 }
}

resources/read example:

json

{
  "jsonrpc": "2.0",
  "id": "4",
  "method": "resources/read",
  "params": { "uri": "fromsrc://docs/installation" }
}

Resource slugs are normalized lowercase kebab-case paths (for example fromsrc://docs/getting-started).

Implementation

Route handler

Create app/llms.txt/route.ts:

typescript

import { getAllDocs, getDoc } from "fromsrc";

export async function GET() {
  const docs = await getAllDocs("docs");
  const sections: string[] = [];

  for (const meta of docs) {
    const slug = meta.slug ? meta.slug.split("/") : [];
    const doc = await getDoc("docs", slug);
    if (doc) {
      sections.push(`# ${doc.title}\n\n${doc.content}`);
    }
  }

  return new Response(sections.join("\n\n---\n\n"), {
    headers: { "Content-Type": "text/plain" },
  });
}

Raw markdown endpoint

Use Next.js proxy to serve .md files:

typescript

// proxy.ts
import { NextRequest, NextResponse } from "next/server";

export function proxy(request: NextRequest) {
  const { pathname } = request.nextUrl;
  const accept = request.headers.get("accept") ?? "";
  if (pathname === "/docs.md" || pathname.endsWith(".md")) {
    const slug = pathname.replace("/docs/", "").replace(".md", "");
    return NextResponse.rewrite(new URL(`/api/raw/${slug}`, request.url));
  }
  if (pathname.startsWith("/docs/") && accept.includes("text/markdown")) {
    const slug = pathname.replace("/docs/", "");
    return NextResponse.rewrite(new URL(`/api/raw/${slug}`, request.url));
  }
  return NextResponse.next();
}

Best practices

Structure content clearly

Use descriptive headings and short paragraphs. AI tools parse structure to understand context.

Include code examples

Practical examples help AI understand usage patterns and generate accurate code.

Keep pages focused

One topic per page makes retrieval more precise and reduces noise in AI responses.