Platform
Specialist agents, one accountable loop.
Everything runs on a schedule against your site. You touch two surfaces: a work queue and a review inbox. The rest is machinery you can audit but never have to operate.
01 · What the platform does
It reads the site before it writes a word.
A crawl of what robots.txt allows becomes a link graph; demand is scored against measurable signals only. Every ranked item in the queue comes out of that reading, never out of a template.
The platform, in sectionThe reading, in order
Crawl, graph, rank, then work.
What a crawl cannot measure is stated as a limitation and lowers the confidence figure; it is never guessed.
02 · The operating loop
Four tokens, one accountable queue.
The loop runs on a schedule. What reaches you is a queue already scored and drafts already checked: each row carries its gate, so you can see at a glance what runs alone and what waits on you.
Two surfaces are yours: the work queue, ranked work collapsed from raw findings, and the review inbox, drafts waiting on your call. The machinery between them is auditable, never something you have to operate.
- 2Surfaces you touch
- 12Agents behind them
- 1Ranked queue
The loop never closes you out: decline reopens a page as new work, and every hand-off is logged back to its evidence.
SearchHandled · running nowSample data
22 items in the queue · collapsed from 60 raw findings
03 · Control & safety
The system works. The call stays yours.
Publishing runs under rules you set, auto, review, or human-only per kind of work, and every hand-off is logged, so any page can be traced back to the evidence behind it.
- 8scoring signals
- 6measured by a crawl
- 3non-negotiables
Operating charterEnforced on every run
Nothing is invented
The scoring model has eight components; a fresh crawl can measure six. We score over what we measured, renormalise, and say so.
Invented inputsMeasured signalsNothing is dropped silently
Every skipped page carries its reason. Hitting the crawl budget is stated as a limitation and lowers the confidence figure.
Silent skipsReason attachedFailure beats fiction
If a run fails we show the reason. We never substitute a demo for a real result. An invented map is worse than an explained failure.
Demo substitutionFailure explained
No inferred evidence · No hidden exclusions · No fictional fallback
See it on your site
The loop starts with a Growth Map.
Enter your site and get an illustrative Growth Map in seconds. The first run shows the queue those agents would start from.