llms.txt vs AGENTS.md, MCP and schema
llms.txt is a Markdown map of a public website for AI tools, while AGENTS.md is a Markdown file of instructions for coding agents inside a code repository. MCP connects AI apps to live tools and data. Schema.org markup describes page facts for search engines. Choose by reader: websites need schema first, repositories need AGENTS.md.
By SeenLens · Last updated
Free, no sign-up. We fetch /llms.txt, /llms-full.txt and up to 10 linked URLs. We do not store the files.
How do llms.txt, AGENTS.md, MCP and schema compare?
The four names come up together because each one gives machines context. They solve different problems for different readers. The table compares them on the points that decide which one you need.
| llms.txt | AGENTS.md | MCP | schema.org | |
|---|---|---|---|---|
| Purpose | Summarize a site and list its key pages | Tell coding agents how to build, test and change a codebase | Let AI apps call tools and read data from other systems | Describe the things on a page (organization, article, product) |
| Who reads it | Some coding agents, audit tools, Lighthouse | Codex, Jules, Cursor, GitHub Copilot coding agent, Gemini CLI, Aider, Zed and other coding agents | MCP clients such as Claude, ChatGPT, VS Code and Cursor | Google, Bing and other search engines |
| Format | Markdown with an H1, a summary and link lists | Plain Markdown, no required fields | A protocol: clients and servers exchange JSON-RPC messages | A vocabulary, usually written as JSON-LD in the page HTML |
| Status | Community proposal (llmstxt.org, 2024) | Open format, stewarded by the Agentic AI Foundation (Linux Foundation) | Open-source standard, also an Agentic AI Foundation project | Community vocabulary founded by Google, Microsoft, Yahoo and Yandex |
| Where it lives | /llms.txt on the website | AGENTS.md at the repository root; nested files in subprojects | An MCP server that you run or host | Inside each page, in a script type="application/ld+json" block |
| Used by Google Search? | No | No | No | Yes, for rich results; not required for AI features |
What is AGENTS.md?
AGENTS.md describes itself as a README for agents. The file sits at the root of a code repository and holds what a coding agent needs and a human README would not: setup commands, test commands, code style and pull request rules. AGENTS.md is plain Markdown with no required fields. In a monorepo, each subproject can have its own AGENTS.md, and the agent reads the nearest file.
## Setup
- Install dependencies: pnpm install
- Start the dev server: pnpm dev
## Tests
- Run pnpm test before every commit.
## Code style
- TypeScript strict mode. No default exports.OpenAI released AGENTS.md in August 2025. In December 2025 the Linux Foundation formed the Agentic AI Foundation with AGENTS.md, Anthropic's Model Context Protocol and Block's goose as founding projects. The AGENTS.md site says more than 60,000 open-source projects use the file.
llms.txt vs AGENTS.md: when to use each
- Use AGENTS.md in any code repository where people use coding agents. It changes how the agent works on your code. It has no effect on your public website.
- Use llms.txt on a public website, mostly a documentation site, to point AI tools at your key pages. It has no effect on agents that work inside your repository.
- Use both if you publish an open-source project: AGENTS.md in the repository for contributors' agents, and llms.txt on the docs site for agents that help your users.
Neither file helps a marketing site appear in ChatGPT or Google AI Overviews. For that, see the guide to whether your site is visible to AI.
llms.txt vs MCP
The MCP documentation defines the Model Context Protocol as an open-source standard for connecting AI applications to external systems. An MCP server exposes tools, data and prompts. An MCP client, such as Claude, ChatGPT, VS Code or Cursor, connects to the server and calls those tools while it works on a task.
llms.txt is a static file that an agent can read. MCP is a live connection that an agent can act through. If you want agents to search your catalog, book a slot or query your API, MCP is the fit. If you want agents to find the right documentation page, llms.txt is enough. Browser agents have a related draft, WebMCP, that registers tools on a web page; see the WebMCP guide.
Sites that run an MCP server can list it in /.well-known/ai-catalog.json so agents can find it. Publish that file only if the server exists; see ai-catalog.json errors.
llms.txt vs schema (JSON-LD)
Schema.org is a shared vocabulary for describing things on a page. Google supports JSON-LD, Microdata and RDFa, and recommends JSON-LD. Schema markup sits inside each page and states facts such as the organization name, the article date or the product price. llms.txt sits in one file and lists pages.
Search engines use schema markup today, for rich results. For AI answers, the evidence is weak for both. Google says structured data is not required for generative AI search and that no special schema.org markup exists for it. An Ahrefs study of 1,885 pages that added JSON-LD found no clear citation gain against matched control pages: ChatGPT +2.2% and AI Mode +2.4%, both not significant, and AI Overviews −4.6%.
If you must pick one for a public website, pick schema markup: it has a confirmed use in Google Search, and llms.txt has none. Start with Organization schema and JSON-LD for each page.
Which one does my site need?
| You run | Start with | Optional extras |
|---|---|---|
| A marketing site or blog | Schema markup (Organization, Article), robots.txt that allows AI search crawlers | llms.txt |
| Developer documentation or an API | llms.txt, Markdown versions of pages | An MCP server for the API |
| A code repository | AGENTS.md | Nested AGENTS.md files per package |
| A product with search, booking or checkout | An MCP server or WebMCP tools | ai-catalog.json that lists the server |
| A shop | Product schema, merchant feeds | llms.txt for help pages |
How to check what your site publishes
The llms.txt checker above fetches /llms.txt and validates it. A full GEO audit checks schema markup, llms.txt, Markdown delivery, /.well-known agent files and WebMCP on one page in a single scan. AGENTS.md lives in your repository, so no website scan can see it.
FAQ
Is AGENTS.md the same as llms.txt?
No. Both are Markdown files for AI, but AGENTS.md lives in a code repository and instructs coding agents, while llms.txt lives on a website and lists its key pages.
Does AGENTS.md help SEO or AI search?
No. AGENTS.md is read by coding agents working in a repository. Search engines and answer engines do not use it to rank or cite a website.
Does MCP replace llms.txt?
They do different jobs. MCP lets an agent call live tools and data. llms.txt is a static reading list. A docs site can offer both.
Is schema better than llms.txt for AI search?
Neither has a proven effect on AI citations. Schema markup has a confirmed use for rich results in Google Search, so it is the better first step for a website.
Who maintains AGENTS.md and MCP?
Both are projects of the Agentic AI Foundation, which the Linux Foundation formed in December 2025.
Sources
- AGENTS.md (format, file location, list of agents that read it, stewardship, adoption; checked 2026-11-06)
- Linux Foundation: Formation of the Agentic AI Foundation (MCP and AGENTS.md as founding projects, 2025-12-09; checked 2026-11-06)
- Model Context Protocol: What is MCP? (definition and supported clients; checked 2026-11-06)
- llmstxt.org: The /llms.txt file (proposal by Jeremy Howard, 2024-09-03; file format; checked 2026-11-06)
- Schema.org: About (community vocabulary founded by Google, Microsoft, Yahoo and Yandex; checked 2026-11-06)
- Google Search Central: Introduction to structured data (JSON-LD, Microdata and RDFa supported; JSON-LD recommended; checked 2026-11-06)
- Google Search Central: Optimizing your website for generative AI features (Google Search does not use llms.txt or similar files; structured data not required; no ideal page length; checked 2026-11-06)
- Ahrefs: Does schema markup increase AI citations? (1,885 pages that added JSON-LD against matched controls, published 2026-05-11; checked 2026-11-06)