~/web/seo

SEOconcept

Everything you publish, measured — and fixed. One website resource, and an agent that runs the loop while you sleep.

Not built yet. This page is the specification, written as the page it will deserve — the ambition and the honesty in one document.

One resource: the website

The app has a single base resource — a website — a node in the tree like everything else, with every view reading from it. A workspace holds as many as you manage.

Where the website lives decides what the app may do. Hosted in Tedo, the connection reads and writes: the app can repair what it finds, because Sites and the CMS share its filesystem. External, added by proving ownership through the Site Verification connector, it reads only: the same eyes, findings as suggestions. That second mode is how an agency runs client sites from Tedo before those clients ever host here.

Findings become fixes

Every other SEO tool ends at the finding. This one hosts your pages, so a finding carries its own repair — applied by the agent through the CMS, inside the scope you granted, logged like everything else. You wake up to a diff, not a report.

Audit finding/blog/context-windows missing meta description
<title>Context windows, and why you never get the whole number</title>
−<meta name="description" content="">
+<meta name="description" content="What a context window really holds — and why you never get the whole number.">
seo-agent applied via CMS · 03:04 · scope ~/web · logged
An audit finding on a hosted page, repaired by the agent overnight — the actor, the scope and the change in one record.

The eight views

Overviewhealth, movers, and what the agent did last night
Search Consolethe connector exists today — impressions, clicks and queries, per property
Sitemapsgenerated and submitted, not audited — we host the sites
Keywordsthe set you care about, grouped by intent, each with an owner
Auditscrawler findings on your pages, each carrying its fix
Page speedCore Web Vitals per page, measured by our own crawler — regressions become findings
Rankingspositions over time — ships later; sourcing SERP data honestly is the hard part ships later
Backlinksgained and lost — the expensive index; last, or partnered ships later

The nightly loop

At 03:00 Automate triggers the audit and our own crawler reads every page. By 03:04 the fixable is fixed — metas, sitemaps, redirects — through the CMS, in scope. At 03:05 what needs a human is filed as work items with owners. At 08:00 you read a diff: what changed, who did it, what moved.

This is configuration, not integration: every operation the app declares is an automation step and an agent tool the moment it exists.

Why this is buildable here

We own a crawler and search engines. Everyone else scrapes Google or buys data; ours runs zoeken.ai, suchen.ai and rechercher.ai — the index is in the building.

We host the sites. Sites, CMS and Static Delivery share the tree — that is what turns observation into action, structurally.

The connectors exist. Search Console and Site Verification are defined in Connect today; the v1 data path is wiring, not invention.

We are the first customer. tedo.ai launches with 110 pages and two catalogues, and every SEO action we take on it by hand is logged — that log is this specification's source.