❓ FAQ – Why sitemap is not enough
Why is a `.well-known/mcp.llmfeed.json` needed?
LLMs and intelligent agents cannot guess your intentions from page content alone. `.well-known/mcp.llmfeed.json` exposes a structured declaration of your service’s identity, capabilities, and terms — in a signed and machine-verifiable format.
Isn’t sitemap.xml enough?
Sitemap only lists URLs — not meanings, not trust, not interaction modes. It works for crawlers, not agents. Agents need purpose, trust levels, APIs, costs, restrictions. That’s what `.llmfeed.json` blocks provide.
Can’t LLMs infer this automatically?
No. They hallucinate. Without clear declarations, they guess APIs, misinterpret business models, and fail to respect limits. This creates security risks, confusion, and agent failure. LLMFeed fixes that.
Why should sites self-declare via `.well-known/`?
The agentic web requires visibility. Self-declaration is the only scalable way for a site to say “here’s what I am, here’s how I work, here’s what’s safe”. It reduces guesswork, hallucinations, and unlocks cooperation with LLMs.
Does this replace SEO?
It replaces part of it. Instead of optimizing for keywords and crawlers, you optimize for clarity and trust. `.llmfeed.json` is the SEO of the agentic era — human- and machine-readable, precise, and verifiable.
How does this relate to what Anthropic calls MCP?
Anthropic promotes the Model Context Protocol (MCP) as a way for LLMs to interact safely with external systems. LLMFeed is a public, open-source implementation of this idea — simpler, file-based, and instantly deployable via `.well-known/`.
What are the advantages of LLMFeed?
It’s lightweight (JSON), modular (by file), compatible with all LLMs, and cryptographically verifiable. There’s no SDK, no infra, no lock-in. Just clarity.
Can any site do it?
Yes. Any static site can host a `.llmfeed.json`. You don’t need a backend. You don’t need OpenAPI. Just a few clear declarations — of metadata, trust, capabilities, and intent.
🧪 Bonus: What do LLMs think of all this?
We asked them. Using a `prompt.llmfeed.json` limited to our compiled `.llmfeed.json` files and the public web, we asked leading LLMs: “What’s the future of SEO, the web, and MCP?” Their answers? Often confused, sometimes poetic, but mostly: they want `.well-known/` clarity too.