As we build the Agentic Web, a confusing alphabet soup of standards is emerging. Three files, in particular, are vying for the attention of modern SEOs: llms.txt, cats.txt, and the new WebMCP protocol.
They often get confused, but they serve three distinct purposes in the lifecycle of an AI interaction. Think of them as Context, Contract, and Capability.
1. LLMS.TXT: The Context (What to Know)
- Role: Documentation for Robots.
- Location: Root directory (
/llms.txt). - Audience: Training crawlers and RAG agents.
llms.txt is essentially a Markdown file that strips away the HTML “cruft” of your website. It provides a clean, token-efficient summary of your content. It answers the question: “What information does this website hold?”
Example:
# My Project
> A concise summary for the LLM.
- [Documentation](https://example.com/docs.md)
- [API Reference](https://example.com/api.md)
2. CATS.TXT: The Contract (Who Can Act)
- Role: Licensing and Permissions.
- Location: Root directory (
/cats.txtor.well-known/cats). - Audience: Compliance bots and legal layers of LLMs.
CATS (Content Authenticity & Tracking Standard) is the legal layer. It replaces the blunt instrument of robots.txt with fine-grained licensing. It answers the question: “Who is allowed to use this data, and for what purpose?”
Example:
License: CC-BY-SA-4.0
Allow: Training
Payment-Endpoint: https://example.com/payment
3. WebMCP: The Capability (How to Act)
- Role: Execution Interface.
- Location: Client-side JavaScript (Runtime).
- Audience: Action-oriented Agents (Browsers, Assistants).
WebMCP is different. It’s not a text file sitting on your server. It’s a live protocol running in the user’s browser. It exposes functions that an agent can call. It answers the question: “What can I DO on this website?”
Example:
agent.registerTool('add_to_cart', { id: '123' });
The Comparison
| Feature | LLMS.TXT | CATS.TXT | WebMCP |
|---|---|---|---|
| Primary Goal | Efficient Reading | Legal Permission | Functional Action |
| Format | Markdown | Key-Value Text | JSON/JS Schema |
| State | Static | Static | Dynamic/Runtime |
| Analogy | The Textbook | The Copyright Page | The Lab Equipment |
How They Work Together
In a perfectly optimized Agentic Application, these three work in harmony:
- The Agent reads CATS.TXT to verify it has permission to be there.
- It consumes LLMS.TXT to understand the domain model and vocabulary.
- It attempts to solve the user’s problem by executing tools exposed via WebMCP.
If you are missing any one of these leg of the stool, your site is either invisible, illegal, or useless to the AI agents of 2026.