🚀 Launching wellknownmcp.org

Make your site LLM-readable, verifiable, and agent-friendly.

We've just launched wellknownmcp.org, a new open specification that lets any website expose machine-readable context, prompts, APIs, and intent — to LLMs, agents, copilots, and voice assistants.

Think of it as

.well-known/

✅ Signed prompts
✅ Declared APIs
✅ Trusted context
✅ Agent-readable capsules

🆕 We invented a MIME type for agents:

.llmfeed.json

We didn’t need a new format — JSON is good.
What we needed was an agreement:
A shared understanding that

.llmfeed.json

• ✅ Flexible
• ✅ Human-readable
• ✅ Open and versioned
• ✅ Works with Claude, ChatGPT, Mistral, open-source models
• ✅ Even interoperable with proprietary internal formats

This is semantic interop, not vendor lock-in.

You can add a

.llmfeed.json

.well-known/

🌐 Why now?

Today, LLMs browse the web like tourists with broken maps.
They guess. They hallucinate. They miss the point.

But what if we gave the web a voice again — for agents?

Instead of scraping, we declare:

• What this domain does
• What actions it exposes
• What content is trustworthy
• What requires credentials
• What you can safely reuse

All inside signed, inspectable

.llmfeed.json

🔍 What you can declare

• 🧠 Prompts → Structured, contextual, signed
• 🔐 APIs → Public or token-based, discoverable by LLMs
• 📦 Exports → Share any page or capsule to an agent in 1 click
• 🧭 Navigation → Feed indexes, trusted flows
• 🧱 Full app interfaces → For mobile, web, voice — declared and signed

No wrapper. No middleware.
Just your intent, clearly declared.

🛠️ Try the tools

• ✅ Prompt Tool (demo)
• 🌐 Ecosystem Explorer
• 📜 The Manifesto
• 📦 GitHub Spec

🧠 Give your agent superpowers

Copy/paste these into your agent or browser 👇

• 🗺 Discovery bundle: wellknown.zip
• 📘 Spec export: spec.llmfeed.json
• 🔍 Site export: wellknownmcp.org.llmfeed.json

Built to be minimal, trustable, and adoptable today.
Simple. Libre. Universal.

💬 We'd love your feedback.
💡 We'd love to see your site join the ecosystem.
🤝 If you're building an agent, this might be your new favorite spec.

🤔 What about the critics?

Yes — we’ve heard the questions:

• Isn't this redundant with OpenAPI or JSON-LD?
• Won’t big LLM vendors just push their own formats?
• Isn’t

  .well-known/

  a fragile vector for something this ambitious?
• Do LLMs even read these files yet?

Fair questions. And here’s the honest answer:

• We don’t think LLMFeed replaces OpenAPI — it adds intent and trust to it.
• We don’t think vendor formats will disappear — but this one’s public, forkable, and inspectable.

• .well-known/

  is not a silver bullet — but it’s where standards start.
• Some LLMs already read

  .llmfeed.json

  — and we’re testing with more every week.

This is not about owning a format.
It’s about building a common surface for meaning, for agents that don’t want to guess.

And even if only 3% of agents support this in 2025 —
that’s more structured understanding than 99% of websites had last year.