# IsAgentReady > AI Agent Readiness Scanner. Methodology version: 2026-09-02. Purpose profiles are the primary interpretation layer. The measured stable-core headline is accompanied by emerging bonus evidence and informational evidence. IsAgentReady helps website owners prepare for AI agents by scanning published technical signals and providing actionable recommendations. Those signals can support automated discovery and interpretation. The scanner does not run agent tasks or predict task success. This is the full context document; for the short index see /llms.txt. The scanner analyzes fetchable initial HTML without executing JavaScript. This is a crawler and non-rendered retrieval signal; it does not simulate every browser agent, which may also use the rendered DOM, accessibility tree, and screenshots. ## Creator IsAgentReady was built by Bart Waardenburg, a Dutch full-stack engineer based in The Hague. It measures stable-core coverage of applicable web signals. Weights and grade bands are versioned policy choices, and static scores do not predict agent task success. - LinkedIn: https://www.linkedin.com/in/bartwaardenburg - Website: https://waardenburg.dev - GitHub: https://github.com/BartWaardenburg ## Pages - Homepage / Scanner: https://isagentready.com - Rankings: https://isagentready.com/rankings - Blog: https://isagentready.com/blog - About / Methodology: https://isagentready.com/about - Developer hub: https://isagentready.com/en/developers - RSS Feed: https://isagentready.com/feed.xml ## Categories and what each one checks ### 1. AI Content Discovery (30%) Whether AI crawlers and agents can reach and discover your content over HTTP. Checks: bot accessibility, robots.txt rules, AI crawler directives, sitemap presence, llms.txt, content negotiation (text/markdown), meta robots tags, content freshness signals, Link response headers (RFC 8288), the Content Signals directive, AGENTS.md, agent content contracts, soft-404 hygiene (unknown paths must return a real 404), and DNS-AID agent records. ### 2. AI Search Signals (20%) Whether your content is machine-parseable for AI search and citation. Checks: JSON-LD structured data, Schema.org types, entity linking, schema validation, FAQPage schema, author attribution, and trust anchor pages (reachable about, contact and privacy pages with real content). ### 3. Content & Semantics (20%) Whether the page structure is understandable to agents and assistive tech. Checks: server-side rendering, heading hierarchy, semantic HTML, ARIA landmarks, image alt text, language attribute, descriptive link texts, form labels, autocomplete, keyboard navigability, document title, button names, aria-hidden misuse, and dynamic widget state. ### 4. Agent Protocols (15%) Published machine-interface signals. Checks include MCP Server Card (SEP-1649), OpenAPI, agents.json, API Catalog (RFC 9727), Agent Skills index, AI Catalog, form quality, interactive surface coverage, live OpenAPI and MCP verification, structured JSON API error responses, RateLimit response headers, and developer documentation linked from the homepage. The A2A Agent Card is emerging bonus evidence and does not affect the headline score. WebMCP is shown separately as experimental, non-scored evidence. A commerce cluster (x402, MPP, UCP, ACP, AP2) is evaluated when commerce signals are present. ### 5. Security & Trust (15%) Published transport and security controls. Checks include HTTPS, HSTS, CSP, security headers, CORS configuration, Referrer-Policy, tool definition security, and auto-submit risk. OAuth Discovery and Protected Resource metadata are evaluated when an auth surface is detected. These static signals do not prove safe task execution. Plus non-scored Training Exposure analysis: Common Crawl presence, training crawler blocking, TDMRep declaration, ai.txt policy, and Web Bot Auth signature directory. ## Scoring - Registered readiness checkpoints are classified as stable-core, emerging bonus, or informational evidence. - Applicability filtering excludes skipped checkpoints and undetected surfaces. - Applicable category weights are renormalized. - Emerging checks are reported as emerging bonus evidence. Experimental checks are informational evidence and do not change the stable-core headline. - Methodology version: 2026-09-02 - Weights and grade bands are explicit versioned policy choices. - Static scores do not predict agent task success. - When stable-core evidence is complete, letter grades use these bands: F (0-19), D (20-39), C (40-69), B (70-79), A (80-94), A+ (95-100) - Score colors: red (0-39), orange (40-69), green (70-100) ## Agent integration endpoints This site implements several agent discovery surfaces that it reports, so it also provides concrete examples. - MCP (Streamable HTTP): https://isagentready.com/mcp - MCP discovery: https://isagentready.com/.well-known/mcp.json - Experimental MCP Server Card: https://isagentready.com/mcp/server-card - A2A Agent Card: https://isagentready.com/.well-known/agent-card.json - Agents directory: https://isagentready.com/.well-known/agents.json - API Catalog (RFC 9727): https://isagentready.com/.well-known/api-catalog - Agent Skills index: https://isagentready.com/.well-known/agent-skills/index.json - OpenAPI spec: https://isagentready.com/openapi.json - JSON API base: https://isagentready.com/api/v1 - Agent guide: https://isagentready.com/AGENTS.md - Authentication: https://isagentready.com/auth.md - Developer hub: https://isagentready.com/en/developers ## Tools - scan_website(url): scan a website for AI agent readiness - get_scan_results(domain): retrieve the latest cached scan for a domain - get_rankings(industry?, grade_range?, limit?): list ranked sites - compare_websites(url_a, url_b): start missing scans and compare methodology-compatible results - get_monitoring_history(site_id, limit?, before?): review past readiness and standards updates - get_monitoring_diff(site_id, run_id): inspect one readiness or standards change - get_monitoring_policy(site_id): read the expected baseline and requirements - get_monitoring_status(site_id): read the latest outcome and required action ## Authentication Public scan, comparison, and ranking tools need no credentials. Owner monitoring reads require the one-time management token in the HTTP Authorization header. Remote MCP monitoring configuration stays read-only; setup and configuration use the owner API or the active WebMCP browser session. See https://isagentready.com/auth.md for details. ## WebMCP browser tools Supported browsers receive the same scan, result, ranking, comparison, and monitoring read tools through `document.modelContext`. The browser surface additionally provides `setup_monitoring`, `update_monitoring_schedule`, `pause_monitoring`, `resume_monitoring`, `verify_monitoring_ownership`, `activate_monitoring`, and `request_pdf_report`. Monitoring changes use the active private browser session and never accept or return management tokens. Pause, resume, ownership verification, and activation require explicit confirmation. PDF requests preserve email verification and rate limiting, and return an async status without a signed download URL. ## Monitoring Monitoring reports both site regressions and methodology changes. A `standards_update` explains newly introduced or revised practices, whether action is required, and which retired checkpoints no longer apply. This keeps owners current when AI agent techniques and best practices evolve, even when their content has not degraded.