# AgentGauge — full reference > AgentGauge scans any website and grades how ready it is for AI agents — whether an agent can discover, understand, trust, and transact with the site. The scan is free and returns an instant letter grade against the AgentReady standard, cross-referenced with Cloudflare's Agent Readiness and specification.website. It reads a site exactly the way an agent does: raw HTTP, no browser. ## What we check AgentGauge grades six areas, mapped to the AgentReady standard and the agent buying journey: - **Discoverability (Find)** — robots.txt AI policy, sitemap.xml, llms.txt, llms-full.txt, HTTP Link headers. If these are missing, agents never see you. - **Content for agents (Evaluate)** — Schema.org JSON-LD, Markdown content negotiation, an /index.md fallback, and whether your content is readable without JavaScript. - **AI Visibility (Recommend)** — will AI answer engines (ChatGPT, Perplexity, Google AI Overviews) cite you? Citable content blocks, answer-shaped pages, author/entity signals, freshness. - **Capabilities (Use tools)** — MCP servers, A2A Agent Cards, OpenAPI specs, and an /agents entry point. - **Identity & access (Trust)** — no bot wall blocking agents, Web Bot Auth, OAuth discovery, and discoverable policies. - **Commerce (Transact)** — agent payment protocols, Product/Offer schema, reachable checkout, and machine-readable pricing. Each check carries a stable AgentReady ID (e.g. AR-DISC-01) and a strength (MUST / SHOULD / MAY). The report leads with a letter grade plus a baseline-conformance badge. ## Products ### Free scan Paste any URL at https://www.agentgauge.ai and get a grade in seconds, with the exact fixes to make your site readable and usable by AI agents. No signup is required to see your grade. ### Watch by AgentGauge — US$99/year Your agent readiness silently decays as you ship: a deploy removes your llms.txt, a CMS change strips your JSON-LD, or a new standard lands. Watch (https://www.agentgauge.ai/watch) re-runs the full scan every week and emails you the moment your grade changes — with the fix — plus a short monthly all-clear otherwise. Billed annually, auto-renews, cancel anytime. Machine-readable pricing: https://www.agentgauge.ai/pricing.md. ### MCP server Agents can run scans directly: AgentGauge exposes a public MCP server (streamable HTTP, no authentication) at https://mcp.agentgauge.ai/mcp with one tool, scan_website. Connect from Claude Code with: claude mcp add --transport http agentgauge https://mcp.agentgauge.ai/mcp. Metadata at /.well-known/mcp.json and /.well-known/mcp/server-card.json. ## Questions ### What is agent readiness? Agent readiness is whether an AI agent can discover, understand, trust, and transact with your website. Agents don't browse like humans: they read raw HTML, follow files like llms.txt and robots.txt, parse structured data, and invoke tools. If those signals are missing, an agent moves on to a competitor that provides them. ### What is AI visibility? AI visibility (sometimes called GEO, generative engine optimization) is whether AI answer engines like ChatGPT, Perplexity, and Google AI Overviews cite you when they answer questions in your space. It depends on citable content structure, answer-shaped pages, author and entity signals, and freshness, which is different from classic search rankings. ### How does the free scan work? Paste your URL and AgentGauge reads your site exactly the way an agent does: raw HTTP requests, no browser rendering. It runs 31 checks across 6 categories, graded against the AgentReady standard with stable IDs and MUST, SHOULD, MAY conformance levels. You see your grade instantly, with no signup required to view it. ### How do I improve my grade? Unlock the full report with your email to see every failing check with the exact fix, from adding an llms.txt and JSON-LD to restructuring content into citable blocks. Most sites can move a full letter grade with a handful of small files and markup changes, then keep it there with weekly monitoring. ## Scoring methodology AgentGauge runs 31 checks across 6 categories. Each check reads the already-fetched homepage HTML or makes a targeted GET, then returns pass / partial / fail / not-applicable with a 0–1 score, the evidence behind it, and (when it isn't a pass) the exact fix. The overall score is a weight-weighted average of every applicable (non-N/A) check, scaled to 0–100 and mapped to a US-style letter grade (A+ at 97, down to F below 60). ### The AgentReady conformance model Checks that map to the AgentReady standard carry a stable ID (e.g. AR-DISC-01) and a conformance strength: - **MUST** — required for baseline agent readiness. - **SHOULD** — strongly recommended; expected unless there's a good reason not to. - **MAY** — optional, additive capability. Each check also has an `applies` scope: **baseline** (relevant to every site) or **conditional** (only counts when the site actually exposes that surface — e.g. OAuth or commerce checks). Conditional checks that can't be proven applicable return **N/A** and are excluded from both the score and conformance. **Baseline conformance** is the headline badge: a site is baseline-conformant only when every applicable baseline MUST check passes. A conditional MUST that is N/A never breaks baseline. The report also tallies met/total for MUST, SHOULD, and MAY across all applicable spec-backed checks. Checks marked "heuristic" below are AgentGauge's own beyond-spec signals — they carry no AR-ID and don't count toward conformance, but they do count toward the score. ### Discoverability (AR-DISC) — Find - **AR-DISC-01 robots.txt with AI policy (MUST, baseline).** Fetches /robots.txt and parses it. Fails if it's missing, if `User-agent: *` disallows the root, or if it blocks any known AI crawler (GPTBot, ChatGPT-User, OAI-SearchBot, ClaudeBot, Claude-Web, anthropic-ai, PerplexityBot, Google-Extended, CCBot, Bytespider, Amazonbot, Applebot-Extended, Meta-ExternalAgent, cohere-ai). Fix: publish a robots.txt that explicitly Allows AI crawlers and points to your sitemap (RFC 9309). - **AR-DISC-02 sitemap.xml (SHOULD, baseline).** GETs /sitemap.xml and confirms it looks like a ``/``. Fix: generate a /sitemap.xml of your canonical URLs and reference it from robots.txt. - **AR-DISC-03 llms.txt (SHOULD, baseline).** Confirms a non-empty /llms.txt exists — a curated markdown entry point for agents. Fix: add an /llms.txt listing your key pages, docs, and policies (see llmstxt.org). - **AR-DISC-04 llms-full.txt (MAY, baseline).** Confirms a non-empty /llms-full.txt — your full docs concatenated as plain markdown for agents that want everything in one request. - **AR-DISC-05 HTTP Link header (MAY, baseline).** Checks for a `Link` response header on the homepage advertising related resources. Fix: emit Link headers (RFC 8288) pointing to your sitemap, llms.txt, manifests, or API. ### Content for agents (AR-CONT) — Evaluate - **AR-CONT-01 Schema.org JSON-LD (SHOULD, baseline).** Parses every `