Agent view — served from /scan.md.
Free scan · no email
A deterministic snapshot of whether agents can discover and read your site. Every result links to the exact file and header we tested, so you can verify it yourself.
We respect robots.txt and Crawl-delay. Our crawler identifies itself as Code4XBot/1.0 and signs its requests. Three scans per day per address.
Scanning
8 of 55 dimensions. The full audit covers discovery, readability, interaction and representation.
rubric v1.4 · snapshot/8
The full report adds interaction, schema accuracy, content structure and representation — plus the generated remediation files for everything that failed here.
One email. No sequence, no newsletter subscription.
Front-end preview. No request left your browser — these results are illustrative and the Python audit engine is not connected yet.
Reading your result
The snapshot covers eight of fifty-five checks, so treat it as a temperature reading rather than a diagnosis. Here is roughly what each band indicates.
AI systems are struggling to read your site at all. Usually one or two structural causes rather than fifty small ones — content that needs JavaScript, or crawlers blocked by a hosting default nobody reviewed.
Typical effort: days to a couple of weeks, and the first fix usually moves the number a lot.
AI can use your site but it is expensive and ambiguous, so you lose to competitors who are easier to read. This is where most well-built sites land, and where the cheapest wins are.
Typical effort: a few days of configuration work, mostly headers, files and structured data.
You are doing better than almost every site we scan. The remaining value is in the interaction layer — whether an assistant can complete your key flows — and in monitoring what engines say about you.
Typical effort: maintenance, plus watching the standards as they shift.
A low score is not a judgement on your website. Almost none of this was a requirement two years ago, and most of it is invisible to every tool you currently run.
What the snapshot covers
Present, and valid against the llmstxt.org spec.
13+ named agents, split by training vs citation intent.
Link tags and HTTP Link headers pointing at your agent files.
Whether you advertise an MCP endpoint at all.
What you return for Accept: text/markdown.
How much of the page survives without JavaScript.
Valid JSON-LD entity types and whether they are linked.
XML accuracy, and whether a markdown variant exists.