/.well-known/mcp Implementation Guide

RFC 8615 compliant automatic MCP server discovery

🚀 Quick Start: 30 Seconds Setup

Get automatic MCP server discovery working in under a minute

Complete Implementation Guide
30-Second Setup: Add Discovery to Your MCP

Step 1: Add Discovery Link

Add one line to your existing /.mcp.json file:

{
  "mcpServers": {
    // ... your existing servers
  },
  
  // ✅ Add this line for automatic discovery
  "llmfeed_extension": "/.well-known/mcp.llmfeed.json"
}

Step 2: Create Enhanced File

Copy your MCP config to /.well-known/mcp.llmfeed.json and add:

{
  "feed_type": "mcp",
  "metadata": {
    "title": "Your MCP Server",
    "description": "Automatic discovery enabled"
  },
  
  // Copy your exact mcpServers config here
  "mcpServers": { /* ... */ }
}
/.well-known/mcp Directory Structure

🔧 Essential Files (RFC 8615)

/.well-known/mcp.llmfeed.json

Main MCP configuration with enhanced discovery metadata

Required
/.well-known/llm-index.llmfeed.json

Semantic sitemap for automatic agent navigation

Recommended
/.well-known/public.pem

Ed25519 public key for cryptographic verification

Optional
Live Demo: See It Working

🌐 Explore Complete Implementation

See a real /.well-known/mcp directory with 24+ files, trust indicators, and interactive preview.

Browse Live Directory

📋 Copy Working Examples

Download working MCP configurations ready for copy-paste implementation.

Automatic Discovery Protocol (Tested June 2025)

✅ How It Works

1
LLM reads your standard .mcp.json if you have one
2
Detects llmfeed_extension link
3
Requests user permission for enhanced features
4
Accesses enhanced metadata and capabilities

Validation: Successfully tested with Claude Desktop - automatic detection with user consent for security.

Trust-Based Autonomy (Future Evolution)

🎯 Trust Levels & Autonomy

🔓 Unsigned: User consent requiredCurrent (2025)
🔏 Signed: Automatic with notificationQ4 2025
🔐 Certified: Fully autonomousQ2 2026

💡 The Breakthrough

LLMFeed's cryptographic trust system solves the core AI autonomy challenge - enabling safe autonomous agent behavior while maintaining security. This is the only viable path from "AI assistant requiring permission" to "autonomous AI agent."

🧠 Pro Tip: Transform any LLM into an expert LLMFeed consultant by feeding it our complete training materials at /train — it becomes instantly knowledgeable about MCP discovery, trust systems, and implementation patterns.

Implementation Roadmap
Phase 1: Basic Discovery
Add discovery link to existing MCP config
✅ Available Now
30 seconds
Phase 2: Enhanced Metadata
Create /.well-known/mcp.llmfeed.json file
✅ Available Now
2 minutes
Phase 3: Cryptographic Trust
Add Ed25519 signatures for automatic discovery
🔄 In Development
Q4 2025
Phase 4: Full Autonomy
LLMCA certification for zero-friction access
🔮 Planned
Q2 2026
Start Your Implementation Now
1

Download Template

2

See Examples

Browse Live Demo
3

Train Your LLM

Expert LLMFeed Consultant
4

Get Support

FAQ & Troubleshooting

📚 Research Project

LLMFeed is experimental research extending MCP. For official docs visit modelcontextprotocol.io. Semi-automatic discovery tested June 2025 with user consent.