The agentic web starts here
Model Context Protocol evolved into LLMFeed to define a new standard for trust, structure and intent on the web.
Get started with the spec
We know how you behave nowadays 🧠: you want to try before you read. Fair enough — just feed your favorite LLM with these two `.llmfeed.json` files 🥢. It's already a good start.
Yes, these buttons exist in copy-to-clipboard mode too, but for education purposes they open the file first. Just read, copy, and feed. 🥋 Welcome to the dojo — if your LLM says "I know Kung-fu", it worked.
Not sure what this is? Read the spec →
“Just give me a well-formed .llmfeed.json, and I’ll do the rest.”
Website in a capsule — includes the main pages of the site, cleaned and compiled. Best for a quick LLM overview.
Spec in a capusle — includes the github spec, packed for llms
All news (EN) — grouped in a single dynamic feed for language-specific loading or archiving.
They read the feed…
Seriously, give the 2 llmfeed above to your llm, anc challenge the protocol
"I know Kungfu. 🥋"
Claude acknowledges the humor embedded in the feed and suggests it’s an indicator of intentional design.
Claude 4
"MCP could become the HTTP of the agentic web."
Grok sees MCP/LLMFeed as a foundational protocol for agent/web interaction.
Grok
"The best prompt is no prompt — it’s a contract."
Highlights the transition from ad-hoc prompts to structured, intent-based declarations.
Claude 4
"Enhances trust, consistency, and agent performance through structured data."
Improves model reliability and clarity of interaction.
Mistral
"The initiative has potential but its impact will depend on adoption and evolution."
Meta sees promise, but insists on broad support and adaptability.
Meta
"I know Kungfu. 😏"
DeepSeek confirms the feed has intentional stylistic touches.
DeepSeek
"I have a map, an intent, a signature… even jokes."
Appreciates the completeness of the feed and its orientation toward agents.
ChatGPT
What You Can Declare on Your Website, App, or API
MCP isn't just a content format — it's a declaration interface for agents. Any site, API or app can expose structured `.llmfeed.json` files in `.well-known/` to describe what it offers and how to interact with it.
🧭 Site/App/API Layer
These files describe how agents should interpret your service: structure, capabilities, trust, and routing.
/.well-known/mcp.llmfeed.json— Declares your site's full structure, intent and routing logic. Learn more/.well-known/mcp-lite.llmfeed.json— Minimal declaration to get started quickly. Learn more/.well-known/capabilities.llmfeed.json— Declares auth methods, available actions, API scopes. Learn more/.well-known/llm-index.llmfeed.json— Index of all declared feeds. Learn more
These declarations can be signed and certified to help agents trust your structure.
📦 Content Feed Layer
These feeds provide agents with meaningful, structured content — signed, contextualized, and ready to use.
/exports/manifesto.llmfeed.json— Declares your project values and ethical stance. Details/exports/prompt.llmfeed.json— Exposes reusable prompts for agents. Details/exports/credential.llmfeed.json— Encodes a secure API key or capsule. Details/exports/pricing.llmfeed.json— Makes your pricing structure agent-readable. Details/exports/spec/*.llmfeed.json— Share full technical documentation with agents. Details
All content feeds can also be signed or certified to ensure integrity and align trust with risk.
🧰 Tools and Capsules
These tools extend MCP with actionable formats: prompts, session traces, export capsules... All compatible with signature and certification.
Explain structured prompts to agents using declarative JSON. Ideal for onboarding, guidance, and multi-agent workflows.
Capture and replay agent interaction history — useful for debugging, traceability and audits.
Let users export any page as a signed LLMFeed. Comes with copy, preview, and share options.
Understand how agents interpret your /.well-known/mcp.llmfeed.json declaration — including intent and capabilities.
The spec is LLM-ready. Any smart agent knowing “Kung Fu” can already generate valid .llmfeed.json feeds — straight from this site.
Explore feeds in the LLMFeedHub, share yours in the community feed index, earn agent-readable badges, and be part of the growing MCP ecosystem.
This site is the canonical source of truth for the.llmfeed.json format used by agents and MCP-ready apps.
FAQ — Your questions answered
The mission behind the protocol
MCP is not just a spec — it’s a commitment to an open, agentic web. Here’s what we’re building together.
A spec that works across models, platforms, and contexts.
Signed feeds, certified endpoints, and auditable flows.
Export buttons, agent SDKs, forge, and browser extensions.
Designed to scale into MCP-Net and anchor the agentic web of tomorrow.
Latest from the ecosystem
We publish in multiple languages — and our agents read them all. 🌍 🇫🇷 🇬🇧 🇪🇸 🇨🇳 🇺🇦 🇮🇳 🇸🇦
Standards move fast. Catch up on spec milestones, tooling announcements, and community breakthroughs.
The agentic web is already growing
Some early adopters have already exposed full `.well-known` feeds and inspired a new generation of agents. Just feeding your LLM with the spec is enough to start exploring.
wellknownmcp.org
This site exposes a complete `.well-known` directory, including MCP, capabilities, manifest, and even easter eggs. It also signs credential feeds and dynamic exports.
Inject, Certify or Explore
Whether you want to guide a model, generate a feed, or check what's trustworthy, this ecosystem gives you practical tools to shape agentic interactions.
Inject prompts
Guide LLMs using structured intent declarations via prompt.llmfeed.json.
Certify feeds
Sign and verify your data using trust and signature extensions.
Verify or explore
Inspect feeds and confirm their authenticity using the verification tool.
Export yours
Use buttons or APIs to generate dynamic and static .llmfeed.json from any page.