# Websites agents can see, and humans sign off on. > sight4agents creates optimized websites and applications with native MCP (Model Context Protocol) support and human-in-the-loop change management, so agents can read and propose changes while people stay in control of what goes live. sight4agents creates optimized websites and applications with native **Model Context Protocol (MCP)** support and **human-in-the-loop change management**. Agents can navigate your content, search it, and propose changes. People review and approve every change before it goes live. - [Get early access](https://sight4agents.com/#early-access) - [How it works](https://sight4agents.com/how-it-works) ## What sight4agents is More and more of the web is read by software agents as well as by people. Most sites are still built only for browsers, so agents have to scrape layouts, guess at structure, and work around scripts. sight4agents builds sites and applications that serve both audiences from the same content: a fast, accessible experience for people, plus structured, machine-readable access for agents through MCP, Markdown, and JSON. When agents help maintain a site, every change they propose goes through a review workflow, so nothing ships without a human saying yes. ## How it works 1. **Build**: Your site or app is generated with semantic HTML, structured data, and clean, fast pages that work for people on any device. 2. **Expose through MCP**: Each site can include a native MCP server, so AI agents can list, read, and search its content through tools instead of scraping pages. 3. **Agents propose**: Agents can draft updates such as new copy, fixes, or new pages. Each one is submitted as a change proposal, never as a direct edit. 4. **Humans approve**: A person reviews each proposal with a clear diff and preview, then approves, requests changes, or rejects it. 5. **Ship with a record**: Approved changes are published, and every step is recorded so you can see who proposed what, who approved it, and roll back if needed. [Read the full workflow](https://sight4agents.com/how-it-works) ## Features - **Native MCP**: A Model Context Protocol endpoint ships with the site, with tools that let agents list pages, fetch content, and search. - **Human-in-the-loop changes**: Agents propose and people approve. Proposals carry diffs, previews, and a reason, so reviews are quick and informed. - **Machine-readable content**: Every page is also available as Markdown, with an `llms.txt` overview, a JSON content index, a sitemap, and structured data. - **Semantic, accessible HTML**: Proper landmarks, a clear heading hierarchy, and readable typography for screen readers, keyboards, and agents alike. - **Fast and lightweight**: Server-rendered pages with minimal dependencies and no client-side framework required to read the content. - **Audit trail and rollback**: A history of every proposed, approved, and published change, so you can trace a decision or undo one. ## Frequently asked questions ### What is sight4agents? sight4agents is a project that creates optimized websites and applications that serve both people and AI agents. It combines native MCP support with a human-in-the-loop change management workflow. ### Why the name? It is a play on words: a **site** with **sight**, for agents. Most websites are built only for human eyes. sight4agents builds sites that agents can actually see, through structured content and MCP, while humans keep the final say. ### What is MCP? The Model Context Protocol (MCP) is an open protocol that lets AI applications connect to external tools and data through a standard interface. A site with an MCP server lets an agent call tools such as "search" or "get page" instead of scraping HTML. ### What does human-in-the-loop change management mean? Agents can suggest changes to a site, but they cannot publish on their own. Each change becomes a proposal that a person reviews and approves or rejects before anything goes live. ### Is the site still good for human visitors? Yes. Pages are responsive, accessible, and fast. The agent-facing formats are generated from the same content, so people and agents always see the same information. ### How can an agent read this website? Start with [/llms.txt](https://sight4agents.com/llms.txt), fetch any page as Markdown by adding `.md` to its path, use [/api/content.json](https://sight4agents.com/api/content.json) or [/search.json](https://sight4agents.com/search.json?q=mcp), or connect to the read-only MCP endpoint at `/mcp`. See [For agents](https://sight4agents.com/for-agents) for details. ### How do I get access? sight4agents is in early development. Use the early access link on this page to get in touch. ## Get early access sight4agents is in early development. If you would like to build a site that works for both people and agents, send a short note about your project. [Email us for early access](mailto:hello@sight4agents.com?subject=sight4agents%20early%20access) --- Source: https://sight4agents.com/. More pages: [How it works](https://sight4agents.com/how-it-works.md), [For agents](https://sight4agents.com/for-agents.md), [About](https://sight4agents.com/about.md), [Brand](https://sight4agents.com/brand.md). Site overview: https://sight4agents.com/llms.txt