Agent-Ready: llms.txt, WebMCP and agent visits

Agent-Ready makes your website easy for AI agents to find, read and use. It works from the same knowledge base as your other agents and has four parts:

• a hosted llms.txt that lists your pages for AI assistants and stays up to date by itself;
• WebMCP tools the ASDK widget offers to AI agents in a visitor's browser;
• a readiness score with the fixes that raise it;
• agent visits: which AI agents fetch your files, load your pages and call your tools.

Agent-Ready is in early access and included in every plan; agent visit analytics come with the Pro plan and higher (see pricing). Open it in the dashboard under Agent-Ready.

Hosted llms.txt

An llms.txt file is a short Markdown list of a site's pages for AI assistants, as proposed at llmstxt.org: a title, a one-line summary, and sections of links, each with a note. ASDK writes yours from your knowledge base:

• the title is your site's name and the summary comes from your home page, unless you write your own;
• every page ASDK has read is listed with a one-line summary. Summaries are written once per version of a page, a few minutes after ASDK reads it; when no summary is available yet, the page is listed with its title;
• pages are grouped by the first part of their address (/docs/… becomes a "Docs" section); blogs, news and similar archives go to the Optional section, which agents may skip when they are short of space;
• key pages you choose come first, and links you add to Optional come last.

There is also an llms-full.txt with the full text of every page and your Q&A pairs, for agents that want everything in one file.

Both files are served from ASDK's API at addresses shown on the Agent-Ready page, such as https://api.asdk.com/v1/sites/SITE_ID/llms.txt. They are cached for five minutes; changes you save in the dashboard show at once. You can switch the files off, for example while you keep an llms.txt of your own.

Serve it at your own address

Agents look for https://yoursite.com/llms.txt. Point that address at the hosted file with a rewrite (your address shows the hosted file) or a redirect (agents follow it to the hosted file). Either way the file stays up to date. The Agent-Ready page shows the code with your addresses filled in for each platform:

For example, on Next.js:

// next.config.js
module.exports = {
  async rewrites() {
    return [
      { source: "/llms.txt", destination: "https://api.asdk.com/v1/sites/SITE_ID/llms.txt" },
      { source: "/llms-full.txt", destination: "https://api.asdk.com/v1/sites/SITE_ID/llms-full.txt" },
    ];
  },
};

Then press Check now. ASDK fetches your /llms.txt and tells you whether it serves the hosted file (through a redirect or a rewrite), serves an llms.txt of your own, answers "not found", or returns something else, such as your home page.

WebMCP tools

WebMCP is a proposed web standard that lets a page register tools AI agents in the browser can call, instead of reading the page and clicking. It is a Draft Community Group Report of the W3C Web Machine Learning Community Group, not a W3C Standard, and Chrome is testing it in an origin trial (Chrome for Developers). Browsers without WebMCP simply ignore the tools.

When the Agent-Ready agent is switched on (on the Agents page) and Offer WebMCP tools is on, the widget registers four tools on every page where it runs:

Each tool has a JSON Schema for its input and returns structured JSON, including errors an agent can act on (for example "try again in 30 seconds"). Following Chrome's guidance on tool security, the three tools that only read are marked read-only, and contact_sales is marked as having real-world effects, so browsers may ask the visitor before it runs. It also refuses to send anything unless the agent confirms that the visitor agreed to be contacted.

Every tool call is counted under Agent visits. The tools are registered with the API described in Chrome's imperative API reference. To try them, turn on WebMCP testing in Chrome as Chrome's WebMCP overview describes and open a page with your widget.

Readiness score

Run the check runs the same checks as the free Agent Readiness Checker against your site: whether AI crawlers may fetch it, whether it has a sitemap and an llms.txt, whether its pages carry their text in the HTML, structured data, and whether agents can act on it. You get a score from 0 to 100 in four categories, the fixes that raise it most with copy-paste code, and the history of your scores.

Agent visits

On the Pro plan and higher, the Agent-Ready page counts per day:

• llms.txt and llms-full.txt fetches by agent: AI search crawlers, assistants fetching for a person, training crawlers, scripts and browsers;
• page loads by AI agents that name themselves, such as the AI user agents of OpenAI, Anthropic and Perplexity, and ChatGPT agent, which identifies its requests with a Signature-Agent header (OpenAI);
• WebMCP tool calls by tool.

Most agentic browsers look like a regular browser to websites, so their page loads are counted only when they name themselves. Their use of your WebMCP tools always counts.