lab — working exhibits
Run the instrument.
Exhibits, not screenshots. The flagship grades any site's readiness for the machines doing the reading — ChatGPT, AI Overviews, Perplexity, agents — live, in full, ungated. Point it at your site. Point it at this one.
09 exhibits installed · scoring model published · agents welcome via mcp
exhibit 01 — agent-ready grader
live engine · ungated · 8 runs/hr per visitor
exh-05 — the readers · the machine guestbook
page reads by known crawlers & agents · rolling 7 days
exh-06 — grader league · every public scan, anonymised
2 scans · rolling 30 days
the bench — what's installed
agent-ready grader
grades any site's readiness for machine readers — engine ported from production, scoring model published
livemcp endpoint
this site as a tool server — agents can call the grader directly
live — POST /api/mcpllms.txt
the machine-readable map of this site, served at /llms.txt
liveconcierge
ask the site about the work — live in the /services/ai-systems terminal, and via mcp as ask_concierge
livethe readers
the machine guestbook — every crawler and agent that reads this site, counted and credited below
live — countinggrader league
every public scan joins an anonymous distribution — domains hashed, grades kept
livemarkdown mirrors
every route served as clean markdown at /md/ — the machine-readable mirror, practised
live — GET /md/vitals loop
real-user web vitals, sampled and aggregated — the footer shows yours, this bench shows everyone's
live — measuringa11y audit
axe-core, wcag 2.2 aa, all eight signed pages — one violation found (hero split aria), fixed same day; zero open
clean — 26 aug 2026mcp: streamable http at /api/mcp — tools: grade_site · site_map · ask_concierge · request_engagement · contact_channel. llms.txt at /llms.txt · markdown mirrors at /md/.