# The agent-readiness package files, explained > Markdown companion to . Same facts as the HTML page. Our agent-readiness files help AI agents that land on your website find and understand your business more efficiently. We do not change your theme, existing plugins, or server code, and we do not add a new backend. The package is 13 files. What goes in it is driven by published standards, working-group proposals, and documented conventions. Every file below links to a plain-English education page describing its origin, who defines or maintains it, and where adoption stands today. Files are grouped by the four Agent Readiness Score dimensions you see in your free Pre-Check and dashboard: Discoverability, Content Understanding, Agentic Capabilities, and Permissions and Governance. ## Discoverability Can an agent find you? These are the files an agent reads first, before it reads a single page. - **`llms.txt`** — proposal-backed, Lighthouse audited. The foundational answer key at your website root. It is a machine-readable overview AI agents read first. Proposed by Jeremy Howard (Answer.AI), now checked by Google Chrome's Lighthouse Agentic Browsing audit. - **`robots.txt`** — standards-backed, IETF RFC 9309. We update your `robots.txt` by adding Cloudflare Content Signals and `Allow: /` for the crawlers behind Amazon, Apple, ChatGPT, Claude, Gemini, Grok, Meta, Perplexity and others. We do not overwrite your existing rules. - **`sitemap.xml`** — standards-backed, Sitemaps.org. The map agents follow to discover all your content. We generate one only when your site has none. An existing sitemap is scored and referenced. We do not overwrite it. - **`.well-known/agent-card.json`** — proposal-backed, A2A (Linux Foundation). A structured capability card in the Agent2Agent protocol format, a standard now backed by 150+ organizations, for static agent-to-agent discovery. ## Content Understanding Does the agent understand your business? These files state what you do in a form a machine can parse. - **`schema.jsonld`** — standards-backed, W3C and Schema.org. Validated Schema.org structured data generated from your actual content and merged with any JSON-LD you already have. Used by more than half of all websites. - **`llm-info.html`** — proposal-backed, emerging convention. A single canonical briefing page for AI systems, following the AI Info Page convention originated by Amin Foroutan. Agents can read one briefing instead of reconstructing your business from fragments. - **`agenticweb.md`** — AgentReadyWebsite convention. The strategic layer: your brand narrative, capabilities, and a recommended agent reading order, in the Markdown format agents parse best. Our own convention, labeled honestly. ## Agentic Capabilities Can an agent act, transact, or use your tools? These files declare what your site exposes. - **`webmcp-registration.js`** — proposal-backed, browser-vendor experimental. An experimental static wrapper for WebMCP, the browser API Google and Microsoft are building so websites can offer structured tools to agents. We ship it and keep the claim scoped to what the file actually does. - **`.well-known/agent-skills/index.json`** — proposal-backed, Agent Skills standard. A static catalog of your business capabilities in the Agent Skills format, the open standard originated by Anthropic and adopted across major AI clients. - **`.well-known/mcp/server-card.json`** — proposal-backed (draft), MCP SEPs. A static Model Context Protocol server card that prepares your site for MCP discovery, without claiming a live MCP server is installed. - **`.well-known/api-catalog`** — standards-backed, IETF RFC 9727. The IETF-standard well-known URI where agents discover your published APIs, or a valid, honest placeholder stating that no public API is advertised. ## Permissions and Governance What are agents allowed to do, and on what terms? These files state your rules. - **`.well-known/agent-permissions.json`** — proposal-backed, LAS-WG (arXiv). A machine-readable permission manifest for AI agents, per the spec published by Oxford, Stanford, and Microsoft researchers. Your rules, published where compliant agents look. - **`.well-known/ucp`** — proposal-backed, Google UCP. Your Universal Commerce Protocol profile, the Google-and-Shopify-led open standard for agentic commerce, or a clear no-commerce placeholder for service businesses. ## Your existing files stay in place Three of these files may already exist on your site. We add to `robots.txt` if you have one. We generate `sitemap.xml` only if you have none. We merge `schema.jsonld` with your existing structured data. We do not silently overwrite a file that is already working. ## Important boundary AgentReadyWebsite delivers static files and installation guidance. We do not modify your theme, existing plugins, or server code, create live server-side MCP/A2A/NLWeb/WebMCP endpoints, host your website, rewrite your site's accessibility tree or layout stability, or guarantee traffic and revenue lift. For WordPress websites, our free installer plugin installs your agent-readiness package from WordPress admin after you upload the delivery zip. No FTP is required. On other platforms, upload the zip. ## How this relates to Google AI Search Google's guidance on AI features in Search says special AI files, AI-only markup, or `llms.txt` are not required to appear in AI Overviews or AI Mode, yet that is what we deliver in our agent-readiness package. The reason we deliver these files is we leave AI Mode placement and SEO rankings alone. We help AI agents that have already reached your website understand your business more efficiently and accurately answer questions about what you offer. See the FAQ at for how generative AI search visibility differs from agentic browsing, which is the job we focus on. ## How this relates to Google's Chrome Lighthouse Agentic Browsing Google Chrome's Lighthouse Agentic Browsing category groups four areas: `llms.txt` presence, WebMCP integration, accessibility tree integrity, and layout stability. The WebMCP area itself has multiple named audits; Lighthouse's expected install pattern is live tool registration on the page, not a static file alone. - **Covered today: `llms.txt`.** Our package includes the root-level `llms.txt` file Chrome's audit looks for. - **Shipped, not yet counted: WebMCP.** We include an experimental static `webmcp-registration.js` wrapper, but do not count WebMCP toward Lighthouse coverage until live registration (declarative form attributes or the browser API) is verified against Chrome's audits. - **Customer-site dependent: accessibility.** Accessibility tree integrity depends on your site's existing HTML, labels, and interactive elements. - **Customer-site dependent: layout stability.** Cumulative Layout Shift depends on your site's CSS, scripts, images, fonts, and rendering behavior. ## See which of these files your site has today The free Pre-Check inventories your agent-readiness files and returns your Agent Readiness Score in seconds: