Agent readiness checkers compared (2026)

What it checks

• 22 checks across 5 categories: Discoverability, Content Accessibility, Bot Access Control, API/Auth/MCP discovery and Commerce. (source, read 2026-10-14)
• Scored checks include an Agent Skills index, WebMCP, API Catalog (RFC 9727), OAuth/OIDC discovery, OAuth protected resource, MCP Server Card, A2A Agent Card, Content Signals, Web Bot Auth and Markdown negotiation. (source, read 2026-10-14)
• Also checks DNS-AID, Auth.md and an ARD manifest. These are not in the launch post's check list. (source, read 2026-10-14)
• Commerce checks (x402, MPP, UCP, ACP, AP2) are shown for information and do not affect the score. (source, read 2026-10-14)
• The A2A Agent Card check is left out of the default All Checks preset and can be added with Customize. (source, read 2026-10-14)
• In our API scan on 2026-10-14 with no check list, the A2A check ran. (source, read 2026-10-14)
• The launch post says it does not check for llms.txt by default. (source, read 2026-10-14)
• The site's own llms.txt lists llms.txt and llms-full.txt under Content Accessibility. (source, read 2026-10-14)

Output

• A score out of 100 and a readiness level from 0 to 5. (source, read 2026-10-14)
• The overall score is passed checks over counted checks across the four non-commerce categories, with one weight per check. (source, read 2026-10-14)
• Each check has a fix prompt and a SKILL.md link, and a Copy all instructions button for coding agents. (source, read 2026-10-14)
• A public scan API answered without a key on 2026-10-14 and can return one combined fix prompt. (source, read 2026-10-14)
• Presets: All Checks, Content Site, API / Application and Customize. (source, read 2026-10-14)

Where it is stronger than w2agent

• Scores an Agent Skills index. w2agent has no Agent Skills check. (source, read 2026-10-14)
• Scores WebMCP, Web Bot Auth and Content Signals. w2agent checks them but does not score them. (source, read 2026-10-14)
• Checks DNS-AID, Auth.md and ARD, which w2agent does not. (source, read 2026-10-14)
• Lists five commerce protocols. w2agent checks UCP on ecommerce sites only. (source, read 2026-10-14)
• Offers a scan_site MCP server at /mcp. w2agent does not offer a public scanning MCP server. (source, read 2026-10-14)
• Registers a scan_site WebMCP tool on its home page. (source, read 2026-10-14)
• Cloudflare Radar publishes adoption data for the top 200k domains, which the launch post says updates weekly. (source, read 2026-10-14)
• The URL Scanner API takes an agentReadiness option, with a Cloudflare account and token. (source, read 2026-10-14)

Limits

• We found no methodology page. (source, read 2026-10-14)
• Result URLs such as /<domain> are filled in by the browser. Fetched without a browser Accept header, one returned 404 with an empty body. With one, it returned the same page shell as the home page. (source, read 2026-10-14)
• Its check list names no structured data or content quality checks. (source, read 2026-10-14)
• Cloudflare's own pages disagree on whether llms.txt is checked (see above). (source, read 2026-10-14)
• The site states: AI-generated recommendations, AI can make mistakes. (source, read 2026-10-14)
• Pricing is not stated. We saw no sign-up step. (source, read 2026-10-14)

What it checks

• 56 core checks in 7 categories, 6 of them tracked at weight 0. Site-type checks (ecommerce, local business) and platform checks (WordPress, Next.js, Shopify, Docusaurus, Nuxt, Astro) are added and scored when detected. (source, read 2026-10-14)
• A2A agent card scored in the default audit and validated against A2A v1.0. A v0.3 card counts as partial. (source, read 2026-10-14)
• MCP discovery runs a live handshake probe and credits an auth-gated endpoint that answers 401 with Bearer resource metadata. The MCP server card is tracked, not scored. (source, read 2026-10-14)
• OAuth discovery follows the issuer-matched RFC 9728 to RFC 8414 chain and adds PKCE S256, CIMD and MCP-endpoint auth checks (11 checks in Auth & Access). (source, read 2026-10-14)
• Structured data (7 checks), content quality (8) and technical hygiene (6). llms.txt gets three scored checks: exists, valid format, and links resolve. (source, read 2026-10-14)
• WebMCP, Content Signals and Web Bot Auth are checked but tracked at weight 0, so they do not change the score. (source, read 2026-10-14)

Output

• Score from 0 to 100 and a letter grade (A at 75 and above, F below 30). Each check ends as pass, partial, fail, skip or error. When the homepage cannot be measured (for example blocked or auth-gated), the audit returns a typed failure and no score. (source, read 2026-10-14)
• Free, no sign-up. Each measured domain has a server-rendered page at /score/<domain>. (source, read 2026-10-14)

Limits

• No Agent Skills check. (source, read 2026-10-14)
• No JavaScript rendering, public surfaces only, and the score is not yet calibrated against real outcomes (the methodology marks calibration as pending). (source, read 2026-10-14)
• WebMCP, Content Signals and Web Bot Auth are tracked but not scored. It has no DNS-AID, Auth.md or ARD check. (source, read 2026-10-14)

What it checks

• Five weighted signals: Discoverable 25, LLM-friendly 15, Catalog 20, Machine-callable API 15, Agent-payable 25. (source, read 2026-10-14)
• MCP detection looks for a card at /.well-known/mcp.json, a JSON-RPC initialize at /mcp and mcp.<domain>, and detects auth-gated 401 and 403 answers. (source, read 2026-10-14)
• Looks for /.well-known/agent.json rather than the A2A agent-card.json path. Checks /llms.txt, Product, Service or Offer JSON-LD, /openapi.json and /.well-known/x402.json. (source, read 2026-10-14)

Output

• Score 0 to 100: 0 to 39 "invisible", 40 to 79 "partial", 80 to 100 "agent-ready". One dataset item per URL. (source, read 2026-10-14)
• Pay per Apify platform usage. (source, read 2026-10-14)

Where it is stronger than w2agent

• Callable through the Apify MCP server and API. w2agent does not offer a public scanning MCP server. (source, read 2026-10-14)
• Checks for an x402 payment file. w2agent does not check x402. (source, read 2026-10-14)

Limits

• Five signals only. No robots.txt, sitemap or bot-access check named. (source, read 2026-10-14)
• Probes agent.json, not the A2A agent-card.json path. Weights are listed, but there is no separate methodology. (source, read 2026-10-14)

What it checks

• Access: robots rules, AI-bot access, JS-rendered content, sitemaps and canonicals. Transaction: WebMCP, structured actions, commerce protocols. Also llms.txt, structured data, canonical and Open Graph tags. (source, read 2026-10-14)
• The full check list loads in a widget we could not read, so this list is from page copy. (source, read 2026-10-14)
• Classifies a site as SaaS, ecommerce, docs or marketing and marks irrelevant checks N/A. (source, read 2026-10-14)

Output

• Three tiers (Access, Citation, Transaction), a score and prioritised fixes, grouped like Lighthouse. The score scale is not stated. (source, read 2026-10-14)
• Free, no sign-up. (source, read 2026-10-14)

Where it is stronger than w2agent

• Names a JS-rendered content check. w2agent does not render JavaScript. (source, read 2026-10-14)
• Names WebMCP checks in its Transaction tier. w2agent tracks WebMCP without scoring it. (source, read 2026-10-14)
• Says plainly where a signal such as llms.txt "has little real-world impact today". (source, read 2026-10-14)

Limits

• The score scale is not stated. (source, read 2026-10-14)
• No per-domain pages, API, MCP server or methodology link stated. (source, read 2026-10-14)

What it checks

• Agent Access: 12 technical checks covering robots.txt, HTTPS, noindex, content in raw HTML, GPTBot, ClaudeBot and PerplexityBot access, llms.txt, XML sitemap references, JSON-LD, sameAs and a freshness date. (source, read 2026-10-14)
• AI Citation: runs 15 ChatGPT prompts (brand, discovery, comparison) and shows whether the brand is mentioned, cited or skipped, and which competitors are named. (source, read 2026-10-14)

Output

• An overall score split into Agent Access and AI Citation, with pass, warning or fail per check. The scale is not stated. (source, read 2026-10-14)
• Free checks without an account; the page showed "3 / 3 free checks left for today". (source, read 2026-10-14)

Where it is stronger than w2agent

• Names noindex and sameAs checks. w2agent has neither. (source, read 2026-10-14)
• Tests what ChatGPT says about the brand with real prompts. w2agent audits the site and does not run prompts against an assistant. (source, read 2026-10-14)

Limits

• No A2A, MCP, OAuth, Agent Skills or WebMCP check. (source, read 2026-10-14)
• No per-domain pages, API or methodology stated. Free use is capped per day. (source, read 2026-10-14)

What it checks

• Measures how agents "discover, parse, and use your documentation". The page does not say docs only, but it ranks API documentation sites. (source, read 2026-10-14)
• Says "22 checks" on the Agent Score page. A company report says "14 of 23" checks. Fern's pages differ, so we say about two dozen. (source, read 2026-10-14)
• Check groups include llms.txt, Markdown, page size, content structure, URL stability, discoverability, observability and authentication. (source, read 2026-10-14)
• The llms.txt checks cover location, the blockquote structure, file size bands, sampled link resolution, links to Markdown and llms.txt directives in HTML and Markdown pages. (source, read 2026-10-14)

Output

• A score from 0 to 100 and a letter grade from A+ ("Exceptional") to F ("Not Agent Supported"). (source, read 2026-10-14)
• A public leaderboard: 247 companies scored and 189 passing (80 and above) on 2026-10-14. Each company has a page and a llms.txt twin. (source, read 2026-10-14)

Where it is stronger than w2agent

• Goes deeper on llms.txt than w2agent's format check: file size bands, blockquote, link-to-Markdown share and llms.txt directives on sampled pages. (source, read 2026-10-14)
• Measures docs content shape, such as page size, content start position, tab serialization and code fences. (source, read 2026-10-14)
• Built on the open-source Agent-Friendly Docs Spec, with an open-source engine at github.com/agent-ecosystem/afdocs. (source, read 2026-10-14)

Limits

• MCP, Agent Skills, A2A and OAuth discovery are not mentioned on the page or in a sample report. (source, read 2026-10-14)
• We did not find a self-serve scanner for your own site. (source, read 2026-10-14)

What it checks

• Access: server-rendered content, bot and firewall access, robots.txt AI stance. Discovery: sitemap, llms.txt, Link headers, structured data. Parseability: Markdown delivery, llms-full.txt, Content signals. (source, read 2026-10-14)

Output

• Scores run 0 to 100: above 80 is reliable, 60 to 79 a solid base, below 60 real gaps. A null result means it could not grade. (source, read 2026-10-14)
• The score is free with no sign-up. The full report with fixes needs a free trial, and plans start at $39/mo billed yearly. (source, read 2026-10-14)

Where it is stronger than w2agent

• Names an llms-full.txt check. w2agent has no llms-full.txt check. (source, read 2026-10-14)
• Names a Link headers check in its Discovery group. w2agent has no general Link header check. (source, read 2026-10-14)

Limits

• No A2A, OAuth, Agent Skills or MCP check. WebMCP is explained but not checked. (source, read 2026-10-14)
• No per-domain pages or methodology stated. (source, read 2026-10-14)

What it checks

• Fetches the page "as six AI crawlers", tests robots.txt "against 34 known AI bots", renders the page in a real browser, and looks at schema, feeds and Markdown surfaces. It follows sitemaps and feeds. (source, read 2026-10-14)
• The rubric is named but not listed or linked. (source, read 2026-10-14)

Output

• A 0 to 100 score in three bands (0 to 40 "Mostly invisible to AI" up to 71 to 100 "Dominant in AI answers"), a five-pillar breakdown and a Level 0 to 4 ladder. (source, read 2026-10-14)
• Level 4 is "Agent-Operable", and the page says the scan does not assess Level 4 surfaces yet. (source, read 2026-10-14)
• Free full scan with no sign-up, rate limited by IP. Paid tracker plans and a report are offered. (source, read 2026-10-14)

Where it is stronger than w2agent

• Fetches as several AI crawlers and renders the page in a real browser. w2agent does not render JavaScript. (source, read 2026-10-14)

Limits

• A2A, OAuth, Agent Skills, WebMCP and MCP checks are not mentioned. (source, read 2026-10-14)
• Results are shown in the browser. No per-domain pages are described. (source, read 2026-10-14)