agent-ready.dev
Scanner operated by Agent Ready (independent).
In short
Scanner that grades a site on 72 checks, 69 mapped to published specs (Vercel's agent readability guide, llmstxt.org and agent protocol specs), with an llms.txt sub-score and a separate MCP server grader. It ranks 4 of 10 on the ten-criterion rubric with 61 of 100, earning most on open method 10 of 10 and dev fit 10 of 10.
What it is
Scanner that grades a site on 72 checks, 69 mapped to published specs (Vercel's agent readability guide, llmstxt.org and agent protocol specs), with an llms.txt sub-score and a separate MCP server grader.
Operator: Agent Ready (independent). Type: Scanner. Home page: agent-ready.dev.
Ten-criterion rubric
| Criterion | Points | What the source says |
|---|---|---|
| Journey: find | 4 of 4 | llms.txt and site checks 1 |
| Journey: read | 4 of 4 | page checks mapped to the Vercel Agent Readability Spec 1 |
| Journey: use | 2 of 4 | MCP and A2A protocol checks reported without score 1 |
| Journey: pay | 1 of 3 | ACP and MPP reported as pre-standard, not scored 1 |
| Checks | 11 of 15 | 72 checks across four categories 2 |
| Real agents | 0 of 15 | does not send an agent to complete a task 2 |
| Open method: catalog | 3 of 3 | each check listed and mapped to a spec 1 |
| Open method: weights | 3 of 3 | overall score is an unweighted pass percentage; the llms.txt sub-score weights are 3x, 1x and 0.5x; the MCP scanner weights are in a table 2 |
| Open method: scale | 2 of 2 | bands: 90 to 100 excellent, 70 to 89 good, 50 to 69 fair, 0 to 49 needs improvement 2 |
| Open method: changes | 2 of 2 | the methodology page carries a last-updated date (2026-09-02) and describes a changed rule for check S5 2 |
| Standards: cites | 3 of 3 | 69 of its 72 checks map to a published spec on the Specs page (96 percent); C7 and C15 have no published document and P25 is not on the map 1 |
| Standards: maintains | 0 of 2 | states that it does not author the standards it grades against 2 |
| Dev fit: api | 3 of 3 | REST API on Pro 3 |
| Dev fit: mcp | 3 of 3 | MCP servers, MIT-licensed 4 |
| Dev fit: cli | 2 of 2 | npm package agent-ready-scanner 4 |
| Dev fit: ci | 2 of 2 | GitHub Action gates pull requests on the score 3 |
| Peers: ranking | 2 of 6 | frozen State of Agent Readability 2026 snapshot 4 |
| Peers: size | 2 of 4 | 2,581 distinct sites scanned between 2026-04-18 and 2026-11-05 in the State of Agent Readability report 5 |
| Fixes: guidance | 4 of 4 | every failing check explains the fix 6 |
| Fixes: brief | 0 of 3 | fix text is per check; no single copy-ready brief or prompt found 6 |
| Fixes: applies | 3 of 3 | generators produce the file: llms.txt, AI robots.txt, agents.json, API catalog and A2A card 6 |
| Repeat scans | 1 of 5 | states that every check is open and reproducible and returns a deterministic pass or fail; no repeat-scan data 2 |
| Free: free | 3 of 3 | anonymous scans with no sign-up 7 |
| Free: limits | 1 of 2 | 3 scans per IP anonymous, 10 per 30 days signed in 7 |
Total 61 of 100, rank 4 of 10. The ranking is on the pick page; the same cells are on the evidence page.
Seven-criterion reading
| Criterion | Score | Evidence |
|---|---|---|
| Coverage | 4 | About 15 of 24 AgentReady identifiers by our mapping; the 72-check list leaves out bot protection, the AI Catalog, WebMCP and CLIs. 2 |
| Transparency | 5 | All 72 check ids, the scoring formula, the llms.txt weights and a dated changelog are public. 2 |
| Reproducibility | not measurable | No per-site lookup was available (the badge read unscored and free scans are 3 per 30 days per IP), so no repeat run could be made without breaking its limits. 10 |
| Access | 4 | Free web scan, markdown mirrors, CLI and a browser extension; the REST API and authenticated MCP sit in the 19 USD Pro plan. 8 |
| Remediation | 4 | A deterministic fix per check, written by the operator rather than generated. 8 |
| Independence of method | 4 | Grades against third-party specs; its main reference, the Vercel guide, is co-written with Ora. 9 |
| Freshness | 2026-10-20 | Latest dated methodology change 2026-10-20; the comparison page was updated 2026-11-04. 2 |
Total of the five numeric criteria: 21 of 25, rank 1 of 6 (tied). Reproducibility and freshness are shown and not summed. Scale 1 to 5, scored by this site on published evidence; reproducibility is a repeat-run result, not a score.
The seven criteria are defined on the method page; this is the reading of 4 November, kept beside the ten-criterion rubric above. Scores are this site's reading of the cited evidence, not the scanner's own claim.
What the sources say
- The methodology lists 72 checks: 15 site checks, 25 page checks, 10 llms.txt checks and 22 protocol checks.2
- The overall score is passing checks divided by total checks; warn counts as fail, and checks made unreliable by a JavaScript-only page are excluded.2
- Rating bands: 90 to 100 excellent, 70 to 89 good, 50 to 69 fair, 0 to 49 needs improvement.2
- A 2026-10-20 change stopped counting an HTML catch-all page as an llms.txt, sitemap.md or AGENTS.md; historical scans were not re-scored.2
- Protocol checks drop out when an endpoint is absent instead of failing the site.2
- A separate MCP server grader (M1 to M13) uses the site's own weights and never affects the site score.2
- The scan is free with no sign-up; a Pro plan at 19 USD a month adds the REST API, an authenticated MCP server and CI gating.8
- Its comparison page describes is-agentic.com as Vercel's scanner, scored by Ora's ranking methodology.8
- It calls itself independent of hosting platforms, CDNs and model vendors.8
Pricing and access
Free scan with no sign-up; Pro 19 USD a month for API, authenticated MCP and CI gating (read 2026-11-04). 8
What it does not do
- Does not test bot protection or what a WAF returns to agents.
- Does not run agents on tasks; it reads files, headers and HTML.
- Does not rank sites on a public leaderboard.
In the Scanner Concordance
agent-ready.dev returned a score for none of the 60 sampled domains on 2026-11-04 (badge says unscored). That is a finding about access: the Concordance only reads scores a scanner makes available without starting a scan its terms forbid. Read the access notes on the Concordance method page.
Where to read next
Positions itself against Ora-scored is-agentic.com; Ora's check list is wider, agent-ready.dev's method is more fully dated.
Sources
- agent-ready.dev/specs. https://agent-ready.dev/specs, read 2026-11-05.
- agent-ready.dev/methodology. https://agent-ready.dev/methodology, read 2026-11-05.
- agent-ready.dev/docs/api. https://agent-ready.dev/docs/api, read 2026-11-05.
- agent-ready.dev/llms.txt. https://agent-ready.dev/llms.txt, read 2026-11-05.
- agent-ready.dev/state-of-agent-readability. https://agent-ready.dev/state-of-agent-readability, read 2026-11-05.
- agent-ready.dev. https://agent-ready.dev/, read 2026-11-05.
- agent-ready.dev/pricing. https://agent-ready.dev/pricing, read 2026-11-05.
- agent-ready.dev vs alternatives (updated 2026-11-04). https://agent-ready.dev/agent-ready-vs-alternatives, read 2026-11-04.
- Vercel: Make your site readable by AI agents (2026-08-28). https://vercel.com/kb/guide/make-your-site-readable-by-ai-agents, read 2026-11-04.
- agent-ready.dev. https://agent-ready.dev/llms-full.txt, read 2026-11-05.
Last reviewed 2026-11-05