# Is Agentic Is Agentic is a free web tool that scores how ready a website is for [[AI Agents]]. You submit a public URL and get back a report with a score, the evidence it observed, and the fixes it recommends. Scans are run and scored by "Ora"; the site says nothing else about who makes it. Why should you care? A growing share of web visits comes from agents acting on someone's behalf: reading docs, comparing offers, filling forms. When an agent can't read a page or recover from an error, it gives up and the visit is lost. Is Agentic turns that vague worry into a list of concrete checks you can work through. ## How the score works - **Essential checks** carry most of the score: server-rendered content ([[Server-Side Rendering (SSR)]]), correct HTTP behavior, clear document structure, recoverable errors, usable controls, and identity metadata (canonical URL, declared language, Open Graph) - **Recommended checks** only apply when the scanner detects a relevant interface: an API, OAuth, GraphQL, an [[Model Context Protocol (MCP)|MCP]] server, a developer portal, or commerce - Checks that don't apply are excluded instead of being counted as failures - Emerging formats earn a bonus and never a penalty I really like that last rule. Agent standards change every few weeks, and punishing a site for skipping last month's proposal would make the score meaningless. Each report also includes an "observed agent journey": one agent's run through the site, with the friction points it hit along the way. It's supporting evidence and isn't scored, but it's the part that explains WHY you got the score you got. The workflow is simple. Fix the failed Essentials first, then the Recommended gaps that apply to your site, then rescan. The report URL stays stable, so you can share it and track progress over time. ## Ways to use it - In the browser - CLI: `npx is-agentic <domain>`, with an optional `--json` flag - Versioned JSON API (`/api/v1/report`), with an [[OpenAPI Specification|OpenAPI]] spec - Read-only MCP server at `https://is-agentic.com/mcp`, with an optional MCP Apps score card - [[Markdown]] through content negotiation (`Accept: text/markdown`, with a `Vary` header) - Discovery files: [[llms.txt convention|llms.txt]], an RFC 9727 api-catalog, `.well-known/ai-catalog.json`, and a `server.json` for the [[Model Context Protocol Registry|MCP Registry]] The site eats its own dog food: every agent-readiness signal it checks, it also exposes itself. That matters. An audit tool that failed its own audit would be hard to trust. ## Limits A report is a point-in-time snapshot of one public URL. It can't see pages behind a login, and a good score doesn't guarantee that agents will succeed on your site. Security, [[Accessibility (a11y)|accessibility]], and privacy still need their own review. ## My take Is Agentic does for agent readiness what [[Lighthouse]] did for [[Web Performance|web performance]]. Once readiness is a number, teams start optimizing for it. That's the strength and the risk at the same time ([[Goodhart's Law]]): the score is a proxy, and the observed agent journey is the closer look at what an agent actually goes through. Worth running against notes.dsebastien.net and the [[DeveloPassion]] sites to see where they stand. ## References - https://is-agentic.com - https://is-agentic.com/mcp - RFC 9727 (api-catalog): https://www.rfc-editor.org/rfc/rfc9727 ## Related - [[IsItAgentReady]] (Cloudflare's agent-readiness scanner) - [[Lighthouse]] - [[AI Agents Web Browsing]] - [[llms.txt convention]] - [[Model Context Protocol (MCP)]] - [[Model Context Protocol Registry]] - [[Generative Engine Optimization (GEO)]] - [[Search Engine Optimization (SEO)]] - [[Web Development (MoC)]] - [[Agent Experience (AX)]]