cd /news/ai-tools/our-docs-scored-67-on-vercel-s-new-a… · home topics ai-tools article
[ARTICLE · art-106633] src=promptcube3.com ↗ pub= topic=ai-tools verified=true sentiment=· neutral

Our docs scored 67 on Vercel's new agent-readiness test

Vercel's new agent-readiness test, is-agentic.com, scored 67 for a production docs site, failing three Essential checks that block headless agents: empty HTML shells on dynamic routes, 200 status codes on error pages, and content that only exists after React hydrates. The scanner, which uses Ora to run 100+ automated checks, groups results into Essential (80 points), Recommended (20), and Bonus (up to 5), and the CLI tool allows piping raw JSON reports into agent loops. The site's observed agent journey revealed auth-protected preview deployments returning 200 with a login shell instead of 401, which was fixed in ten minutes by adding a proper status header.

read2 min views5 publishedAug 21, 2026
Our docs scored 67 on Vercel's new agent-readiness test
Image: Promptcube3 (auto-discovered)

npx is-agentic

against our production docs site last night. Came back 67 — Essential layer failing on three checks that literally stop an agent cold: empty HTML shells on dynamic routes, 200 status codes on error pages, and content that only exists after React hydrates. The scanner doesn't care about your PageSpeed score. It cares whether a headless client can actually read the page.Vercel launched is-agentic.com last week. Ora does the heavy lifting — 100+ automated checks, visualizations of an agent walking your site, one-click fix prompts. Vercel groups results into Essential (80 points), Recommended (20), and Bonus (up to 5). Smart part: checks only fire when the relevant surface exists. Our commerce endpoints don't get penalized for missing MCP manifests because we don't ship MCP. API and GraphQL checks only trigger when the scanner finds those endpoints. Duplicate check IDs across MCP surfaces get averaged so you're not double-dinged for the same gap.

The CLI is what made this practical for us:

npx is-agentic docs.ourcompany.com --json

Returns the stored report instantly if cached, otherwise kicks off a scan and waits. --json

gives you the raw API body — structured errors, RFC 9457 problem+json format — which you can pipe straight into an agent loop. No HTML scraping needed. Public pages also serve markdown via Accept: text/markdown

with proper Vary: Accept

, so crawlers get clean content without executing JS. The HTML report is complete in the first response too.

Reports cache for ~6 hours. Re-scanning right after deploy hands you the old number. Learned that the hard way — waited for the snapshot timestamp to move before trusting the new score.

What changed our workflow: the observed agent journey. One agent, one walk-through, shows exactly where it got stuck. Not part of the numeric score, but the clip revealed our auth-protected preview deployments returning 200 with a login shell instead of 401. Agents choke on that. Fixed it in ten minutes by adding a proper status header.

The site dogfoods itself — is-agentic.com

is agentic. Your agent can call the API (GET https://is-agentic.com/api/v1/report?url=...

, no auth) to read its own score and close the loop. Chris from Vercel posted the same day: "Best part: your agent can use it to make your site more agentic." That's not marketing fluff; the stack matches the checklist they grade you on.

We're wiring the CLI into our preview deploy pipeline now. Fail the build if Essential drops below 70. Recommended items become PR comments. Bonus stays aspirational. If you're shipping anything agents might touch — docs, APIs, MCP servers, commerce — run the scan. The number matters less than the failed Essential items. Those are the hard stops.

Next Limited AI tooling forced our team to master prompting →

these AI tool field notes, with plenty of directly applicable cases.

── more in #ai-tools 4 stories · sorted by recency
── more on @vercel 3 stories trending now
sponsored brought to you by zahid.host 4,200+ EU-deployed projects
reading about agents? ship yours in a single git push.

Run your AI side-project on zahid.host

EU-based hosting, git-push deploys, automatic HTTPS, no cold starts. Free tier with a custom domain — perfect for shipping the agent you just read about.

$git push zahid main
Live at https://your-agent.zahid.host
Get free account → Pricing
from €0/mo · no card required
LIVE [news/our-docs-scored-67-o…] indexed:0 read:2min 2026-08-21 ·