§4.1 · TOOL · LAST REVIEWED SEPTEMBER 2026
ora.ai agent readiness: Scan, Journey and the ax CLI explained
A testing platform for the agent experience: can an agent discover your site, read it, use it and pay you.
Agent Readiness Compare editors · Spec status checked September 2026
Answer
ora.ai tests how ready a website is for AI agents. Its Scan scores public URLs across four layers (discovery, access, usability and payments) with 119 published checks, and Journey records a real agent attempting a chosen task and shows where it stalled. It is built for developer workflows (CLI, REST API, MCP server), and it published AgentReady, an MIT-licensed specification of what to implement. Watch-outs: new users are asked to join a waitlist, and it tests your stack rather than hosting fixes for you.
On this page
1.What does ora.ai do?
ora.ai describes itself as 'the standard for optimizing your site so agents can actually use and recommend you,' and its home page asks 'Is your site ready for agents?' It scans a domain or an MCP server URL, returns an agent readiness score with fixes ranked by impact, and lets you watch an agent try the site.
2.How does ora.ai Scan score a site?
| Layer | Question (ora.ai wording) | Points | Checks |
|---|---|---|---|
| Discovery | Can an agent find and recommend you? | 20 | 16 |
| Access | Can it access your data and understand you? | 30 | 41 |
| Usability | Can it use you: operate, authenticate, integrate? | 40 | 56 |
| Payments | Can it pay you? | 10 | 6 |
| Total | 100 | 119 |
| Grade | Score | Band |
|---|---|---|
| A+ | 95–100 | Leading |
| A | 86–94 | Agent-Ready |
| B | 70–85 | Competitive |
| C | 48–69 | Needs Work |
| D | 28–47 | At Risk |
| F | 0–27 | Unusable |
ora.ai states that 'only checks we have verified agents rely on count,' and that every score is computed from public URLs, 'so it reflects what a product actually exposes to agents, not what it claims.' Scans run against public URLs only and cannot evaluate private, internal or login-gated content. A fresh scan takes about 30 seconds, with deeper checks resolving afterwards.
Source: ora.ai/methodology · ora.ai/llms.txt · Reviewed Sep 2026
3.What is Journey?
Journey points a real agent at a page with an intent you choose and records its path: 'what it found, where it stalled, whether it finished.' It is the part of ora.ai that tests tasks rather than files, such as whether an agent can find a price or reach a sign-up step. ora.ai's Deep Scan v2 (July 7, 2026) describes the same approach in its scoring: 'we ran the agents, then scored what they did.'
Source: journey.ora.ai · ora.ai blog, Deep Scan v2 · Reviewed Sep 2026
4.How do developers use it?
- CLI:
npx ax auditscores any URL and prints what to fix. - MCP server: add the ora.ai MCP server (https://ora.ai/api/mcp) to an MCP-capable agent and ask it to scan a domain; it reads every failing check and can rescan after each fix.
- REST API for scans and score lookups, and human-readable reports at ora.ai/score/{domain}.
- WebMCP audit: loads a site in a real browser and scores the tools it gives an agent out of 100 (shared experience, tool selection, tool quality, trust). Local run:
npx ax webmcp-audit http://localhost:3000, no sign-up.
npx ax audit https://example.comSource: ora.ai home page · ora.ai/docs · ora.ai/llms.txt · webmcp.ora.ai · Reviewed Sep 2026
5.What is AgentReady?
AgentReady is an open specification ora.ai published on March 20, 2026 (v1.0.0 stabilised April 24, 2026) that 'names what to implement' for agent readiness. It is MIT-licensed, published at agentready.org with a machine-readable version at agentready.org/spec.json, and has five sections: discoverability (robots.txt, llms.txt, sitemaps), content for agents (JSON-LD, markdown negotiation), capabilities (MCP, OpenAPI, WebMCP, NLWeb), identity and access (OAuth 2.0, Web Bot Auth) and commerce (x402, ACP, UCP, MPP). ora.ai describes it as a specification, not a scoring system.
Source: ora.ai blog, AgentReady · agentready.org · Reviewed Sep 2026
6.What is the partnership with Vercel?
On August 20, 2026 ora.ai announced that it partnered with Vercel to launch is-agentic.com: 'enter a URL, get a score and the fixes to raise it.' Every scan is run by ora.ai, and the checks are tuned for websites: essential checks (80 points), recommended capabilities (20 points) and bonus signals (up to 5 points). is-agentic.com's own footer reads 'Powered by Ora.'
Source: ora.ai blog, is-agentic with Vercel · is-agentic.com · Reviewed Sep 2026
7.What does ora.ai cost?
ora.ai states that its free tier requires no API key, sign-up or credit card. Paid plans and prices are not published. The home page invites new users to join a waitlist.
Source: ora.ai/llms.txt · ora.ai home page · Reviewed Sep 2026
8.Where is ora.ai strong, and what should you watch?
Strong
- Tests tasks, not only files: Journey shows where a real agent stalls.
- Most published scoring method of the scanners we reviewed: layers, points, check counts and grade bands.
- Built for developer workflows: CLI, REST API, MCP server and a local WebMCP audit.
- Covers all four layers from discovery to payments, and publishes the AgentReady specification.
Watch
- It tests; it does not host fixes. Changes are made on your own site, CDN or API. Cloudflare scores higher on fix path.
- Access: the home page asks new users to join a waitlist. Cloudflare scores higher on availability.
- Public URLs only: login-gated flows are outside what a scan evaluates.
- Paid pricing is not published.
9.How does ora.ai score here?
Scanner score (editorial assessment, 0-5). Compares readiness scanners on what their public pages document. It is not a measure of overall product quality.
| Criterion | ora.ai |
|---|---|
| Standards coverage (20%) | 4.5 / 5AgentReady spec covers robots.txt, llms.txt, sitemaps, JSON-LD, markdown negotiation, MCP, OpenAPI, WebMCP, NLWeb, OAuth 2.0, Web Bot Auth, x402, ACP, UCP and MPP. |
| Real-agent task testing (20%) | 4.5 / 5Journey records a real agent attempting a chosen intent and shows where it stalled; Deep Scan v2 scores what agents did. |
| Published methodology (15%) | 4.5 / 5Methodology publishes four layers, points per layer (20/30/40/10), check counts (16/41/56/6) and grade bands A+ to F. |
| Fix path on the same platform (15%) | 3.5 / 5Fixes are ranked by impact and the MCP server can rescan after each fix; implementation happens on your own stack. |
| General availability (10%) | 3.0 / 5Free tier needs no key or sign-up; the home page asks new users to join a waitlist. |
| Developer workflow (20%) | 4.5 / 5npx ax audit CLI, REST API, MCP server and a local WebMCP audit (npx ax webmcp-audit). |
| Weighted total | 4.2 / 5 |
| Cloudflare Is It Agent Ready weighted total (for context) | 3.3 / 5 |
ora.ai leads on: Real-agent task testing, Published methodology, Developer workflow. Cloudflare leads on: Fix path on the same platform, General availability. Tie: Standards coverage.
Source: ora.ai methodology · ora.ai llms.txt · ora.ai AgentReady post · journey.ora.ai · webmcp.ora.ai · isitagentready.com · Cloudflare Markdown for Agents · Cloudflare AI Crawl Control · Cloudflare Web Bot Auth · Reviewed Sep 2026
Frequently asked questions
What does ora.ai check?
Four layers: discovery (16 checks), access (41), usability (56) and payments (6), totalling 100 points and graded A+ to F, according to ora.ai's methodology page.
Is ora.ai an AEO or GEO tool?
No. ora.ai tests whether agents can discover, read, use and pay a site. AEO and GEO tools measure how often a brand appears in AI answers.
How is ora.ai related to is-agentic.com?
is-agentic.com was launched by Vercel and ora.ai together, and ora.ai runs every scan.