AI Protocols (llms.txt, NLWeb, MCP)
Emerging standards that make your site natively accessible to AI agents.
What We Test
AgentReady checks for five AI-native protocols: llms.txt (a plain-text file that describes your site to language models), NLWeb (a .well-known endpoint for natural language queries), MCP (Model Context Protocol configuration at .well-known/mcp.json), AI Sitemap (an AI-specific sitemap at /ai-sitemap.xml), and structured APIs (OpenAPI/Swagger specs). These are the emerging standards for the AI-native web.
Why It Matters
The web is transitioning from being built for browsers to being built for AI agents. Just as robots.txt told search engines how to crawl your site in the early 2000s, llms.txt tells AI models what your site is about and how to use it. NLWeb lets AI agents query your site with natural language instead of keyword searches. MCP allows AI assistants to directly interact with your site's data and APIs. Sites that adopt these protocols early will have a significant advantage as AI search becomes dominant.
How Scoring Works
AI Protocols account for 15% of your overall score. Points: llms.txt presence earns +25, NLWeb endpoint earns +30 (the highest single item — reflecting its importance), MCP configuration earns +25, AI sitemap earns +10, and structured API (OpenAPI/Swagger) earns +10.
How to Improve
Start with llms.txt — create a file at yourdomain.com/llms.txt that describes your site, its purpose, key pages, and how AI should interact with your content. This is the simplest and highest-impact protocol to implement. For NLWeb and MCP, AgentReady's Pro plan can auto-generate and deploy these endpoints for your site. An AI sitemap is a lightweight XML file listing your most AI-relevant pages. If you have an API, publish an OpenAPI spec to make it discoverable by AI agents.