Vercel Introduces ‘Is Agentic’, a Free Agent-Readiness Scoring Tool That Audits Public Websites Using Ora’s 100+ Checks
MarkTechPost Michal Sutter
Vercel launched Is Agentic, a free tool that scores how easy a site is for AI agents to use. It runs 100+ checks, and the odd part is that marketing sites aren’t punished for missing features they never promised.
Based on reporting by MarkTechPost, Michal Sutter — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
Vercel has released Is Agentic, a public scoring tool that judges how well a website can be discovered, accessed, understood, and used by AI agents. The scans themselves are run and scored by Ora, the agent-experience research company from era labs. Vercel handles the interface, report pages, storage, and the grouping behind the score.
The tool is free right now, with no paid tier, subscription fee, or per-report charge. No API key or billing account is needed for the public site, the read-only API, the CLI, or the MCP server. A URL can be checked in the browser, and the CLI can also be run from the command line.
Ora’s method breaks the audit into four layers: Discovery gets 20 points across 15 checks, Access gets 30 points across 41 checks, Usability gets 40 points across 56 checks, and Payments gets 10 points across 6 checks. That adds up to 118 checks, which fits Vercel’s claim of “100+ checks.” Ora says the checklist comes from real agent runs, not from someone’s opinion about what should matter. It also publishes letter grades from A+ down to F.
Vercel then reshapes those checks into its own scoring model. Essential checks share an 80-point pool, Recommended checks share 20 points, and Emerging signals can add up to five bonus points without ever lowering a score if they’re missing. Checks that don’t apply are excluded, and partial results get partial credit. A marketing site, for example, is not dinged for failing to expose an API or a commerce flow it never claimed to have.
The reports are meant to be useful, not just decorative. Each finding includes evidence and, when available, a recommendation. The live API response includes fields like score, score_label, scanned_at, eligible_checks, a score_breakdown, and an issues array, with issue details such as id, name, recommendation, result, and tier. Vercel also shows an observed agent journey, but keeps that out of the numeric score because one journey can’t stand in for every agent.
The machine-facing side is unusually polished. The report API is rate-limited to 120 requests per client IP every 60 seconds, errors use RFC 9457 problem details, and the OpenAPI description is the supported integration contract. The MCP server exposes methods for fetching a report, the methodology, and developer docs. And the site itself serves the score in initial HTML, offers a Markdown variant, and sets Vary: Accept so caches don’t mix them up.
My take — AI-written commentary, not fact-checked reporting
This is the rare scoring tool that seems built by people who actually expect websites to be used by machines, not admired in a demo. The clever bit is the restraint: missing APIs don’t get punished on a site that never claimed to have one, which is a refreshingly adult move in a market full of self-congratulatory rubrics. More of this, less fake objectivity with a glossy badge on top.
Read more about this at: MarkTechPost
