🧵 Thread — What if your site could talk to LLMs?

1/
🚀 We just launched https://wellknownmcp.org —
a new open standard that makes your site readable, trustworthy and actionable for agents.
Think

.well-known/

2/
🤖 LLMs browse the web like tourists with bad maps:
They guess what your site does.
They hallucinate APIs.
They miss your intent.

It’s time for clarity.

3/
With a few signed

.llmfeed.json

• What can an agent do here?
• Are prompts trusted?
• Can I act or fetch context?

It’s like a sitemap — but for agent logic.

4/
📂 Key feeds you can expose:

• mcp.llmfeed.json

  → root declaration

• capabilities.llmfeed.json

  → exposed tools or APIs

• prompts/

  → signed prompt capsules

• llm-index.llmfeed.json

  → bundle index for agents

5/
✅ Real use cases:

• Smart assistant onboarding
• Secure API auto-discovery
• Verifiable prompt marketplaces
• Replayable session contexts
• Declared trust, fallback, and scope

6/
🧪 Try it live:

• 🛠 Prompt Tool
• 🌐 Ecosystem Explorer
• 📜 Manifesto

No login. No vendor lock. Just clarity.

7/
💡 What is

.llmfeed.json

A dedicated MIME-type for LLMs:

• JSON-based
• Human-readable
• Machine-optimised
• Open & extensible

It works with ChatGPT, Claude, Mistral, or any custom agent.

8/
💬 Try this with your favorite LLM:

“Here’s a

llmfeed.json

. Explain what this site enables.”
“Show me which prompts or APIs are exposed.”
“How would you interact with this domain?”

Yes — you can turn any LLM into a protocol teacher.

9/
Ready to explore?

• Expose a few feeds
• Sign or certify them
• Submit to the ecosystem

📦 All open-source
📜 Spec: github.com/wellknownmcp

—

🧵 Fin
We're building the missing bridge between the web and agents.
Help shape it. Implement it. Spread the spec.

#LLM #AI #Web #PromptEngineering #SemanticWeb #MCP #LLMFeed