Product

Changelog

What shipped in Dokeo - product updates for the content quality gate.

Google

01 · Draft

Paste, PR, webhook, or CLI. The gate does not care who wrote it.

What shipped on Dokeo, in order. We're a small team building in public beta - this is the real history, not a roadmap.

2026-08-06

IMPROVED5-tier pricing (Free, Growth, Pro, Scale, Enterprise), matching what's on /pricing. Scale is now self-serve checkout at $499/mo.
FIXEDScale tier's scan and Copilot allowances were marked 'unlimited' but had no real ceiling - a single tenant could saturate gate compute indefinitely. Capped to 200,000 scan credits/mo (30k interactive + 170k CI) and 1,000 Copilot queries/mo. Anything bigger is Enterprise (contact sales, manually provisioned). (Limits later revised downward - see /pricing for current numbers.)

NEWDokeo Copilot (beta): grounded Q&A over your own scan, citation, and regression data. Web at /copilot plus an ask_dokeo MCP tool. One constrained GLM call phrases a deterministic context; if the data isn't there, it says so instead of guessing. Credit-metered separately from scans and rewrites. See /pricing for current allowances per tier.
NEWDokeo Connectors (beta): bring-your-own-key data import framework. First connector is Ahrefs (Site Explorer metrics over their v3 API); Semrush, Moz, Voluum, and Keitaro follow the same interface. Tenant API keys are Fernet-encrypted at rest, never proxied by Dokeo, and rate-limit errors surface verbatim so tenants see their own quota.

FIXEDSecurity: closed an unauthenticated admin-header spoof on the growth dashboard's admin check (the Fly app is directly internet-reachable, so anyone could set X-Dokeo-Is-Admin: 1 and read prospect data). Identity now comes from the verified session JWT, never the unsigned header.
FIXEDCredit-ledger race that could lose concurrent updates - now locked across processes from day one, with a regression test.
NEWRecruiter-side screening pipeline with bidirectional matcher (Tarkflo-parity).
NEWRole dashboard renders HTML instead of JSON, with a free TTS fallback chain.
IMPROVEDReplaced 8 stub pages with real functional UIs across the app.

NEWOne-click AI rewrites are live on Growth+ plans, powered by GLM 5.2. Rewrites consume from the separate rewrite allowance, never from scans, and the scoring path stays fully deterministic.

NEWScan a live URL from the homepage demo - drop a link and we fetch and gate the article, no pasting or signup needed.
NEWCustom checks: add your own pass/fail rules (required phrases, banned words, headings, word or sentence limits) that run alongside the built-in checks.
IMPROVEDScan results now show a per-check breakdown - which checks passed, flagged, or failed and why - not just the score.
IMPROVEDContent generation is faster and more reliable under load; rate-limit failures now retry cleanly with a clear message instead of a raw error.
IMPROVEDResponsive and accessibility pass across the app: two-column Scan and Builder layouts on desktop, tables that scroll on mobile, a slimmer cookie bar, and form labels wired for keyboard and screen readers.
FIXEDQuick tour: step icons, accurate copy, a personalized final step, and completion now records correctly.

NEW/security and /changelog pages.
NEWllms.txt published for AI crawlers, linked in the footer.
NEWFAQPage structured data on the homepage for rich-result eligibility.
NEWAuto-generated Open Graph / Twitter share image sitewide.
IMPROVEDAbout page: founder section now says Founder (was an unexplained Co-founder) with real links.
IMPROVEDClarified AI rewrites status across the site: one-click AI rewrites are coming soon. Today the gate shows the exact reason each check failed so you can fix it, and every scan feeds the scan-to-outcome dataset.

IMPROVEDPer-page metadata, sitemap, robots.txt, custom 404, and JSON-LD across the site.

NEWAbout page redesign and a 3-minute onboarding tour.
NEWHomepage demo scan widget - try a scan with no signup.
NEWRazorpay Standard Checkout for pricing upgrades.
IMPROVEDNav split between the logged-in app and the logged-out marketing site.
IMPROVEDIntegrations list and surface count trimmed to match what's actually built.
IMPROVEDBatch scan now accepts an explicit URL list, not just a domain crawl.
IMPROVEDScanner shows all 17 content types and 4 builder formats with multi-LLM provider selection.
FIXEDRSS auto-discovery and retry logic for transient fetch failures.
FIXEDai_output content type scans that were crashing on a missing severity mapping.

NEWPublic beta launch: MCP server, custom rules, multi-tenant, credit system with 2x carry-forward, bring-your-own-key builder.

NEWGEO scoring and multi-engine support - 3 engines, 17 content types, a full regression suite.

NEWAEO (answer-engine) checks, after ChatGPT started eating organic search traffic.