A marketing engine that reads real numbers off your own websites every night, writes finished social posts from those numbers, and publishes them for you — and simply holds back any post it cannot back up with a fact.
Why this is different
Most "AI marketing" tools do one thing: they generate plausible-sounding text on demand. That text reads fine and is often quietly wrong — a number no one checked, a claim no one can source, a milestone that never happened. It looks like marketing and behaves like noise. This system was built to be the opposite of that in four concrete ways.
It runs on real data, with an honesty guard
Every post is assembled from verified fields — a site's real description, a real marketing principle from the skills library, a real week-over-week improvement in an AI-readiness score. The templates are fixed sentence patterns that fill those fields in. If a required field is missing, the post is held, not filled with a guess. That single rule is why it is safe to point at a client's brand and let it run.
Flat-subscription economics
The content engine uses deterministic templates, not a per-post AI call. Generating one post and generating a thousand cost exactly the same: nothing. The whole system runs on a flat $24/month for the publishing server. Your cost does not climb as your output does — a structural advantage no pay-per-token tool can match at scale.
A human-approval option when you want it
Each tenant can run fully automatic or route every post through a private, phone-friendly approval queue where a person taps Approve, Edit, or Skip. The owner's own network runs auto-fire; a nervous new client can start with approval on and turn it off once they trust it. Same engine, your call.
It closes the loop
It does not just hand you drafts and walk away. It gathers, drafts, publishes, and then reports — a complete nightly cycle from source data to a live post to a digest of what went out. Publish, then measure. Not a draft mill.
The short version. Real data instead of invention, a guard that refuses to guess, a cost that does not scale with output, optional human sign-off, and a loop that actually ships. That is the whole pitch, and every one of those is a fact you can check in this manual.
The self-driving loop
You do not run this engine by hand each day. It runs itself on a nightly schedule. Here is the whole cycle in plain English — the same four stages happen every night, in order, without anyone touching it.
05:00 — gather
Read the real data. On the main server, the feed builder scans the website network, pulls the verified facts for the night — site features, marketing principles, score improvements — and ships that feed over to the publishing hub.
↓
05:10 — draft
Write finished posts. On the hub, the generator matches each feed item to a template. If every required field is present, it writes a real, ready-to-send post into the queue. If a field is missing, it holds the post instead of guessing.
↓
every 10 min — publish
Send what is due. Around the clock, the publisher checks the queue and sends anything whose time has come — respecting the paid-channel cap and the kill switch. Auto-fire tenants go straight out; approval tenants wait for a human tap.
↓
05:20 — report
Say what happened. The reporter writes a digest of what went out and what was held, so you can glance at the night's work in the morning instead of babysitting it.
A nightly backup runs quietly at 04:17. If anything in the chain fails, the system pushes an alert to a phone channel rather than failing silently. For the full architecture behind these four steps — the two servers, the containers, and the cron chain — see How It Works.
Who this is for — and how to read it
This manual is written for three kinds of reader, and you can pick your own path through it.
The owner / operator
If you run the system day to day, this is your reference for everything: how the pieces fit, how to connect a channel, how to add a client, and what to do when something breaks. Read How It Works and Daily Operations first, keep Playbooks open as your task manual, and bookmark Troubleshooting.
A partner
If you are evaluating this as a business, start right here on this page, then read Why this is different above and the Managing Clients page. Together they show the model, the economics, and how it is sold as a service.
Clients
If you are a client whose brand this engine posts for, the pages you care about are The Content Engine (what it says and why it is trustworthy) and Security, Privacy & Compliance (how your accounts and data are protected). Everything runs isolated to your tenant.
Suggested first read. New to the whole thing? Read this Overview end to end, then
How It Works, then skim the
Content Engine. That is roughly twenty minutes and you will understand the entire system. Everything after that is reference you reach for as needed.
Five-minute quick start
Want to see it real before you read another word? Do these five things in order.
- See it running. Open postiz.wholetech.com — the live publishing hub — and look at the accounts it is posting to right now: X at @springnet and Bluesky. The first live auto-post was verified going to both. This is not a mockup; it is the system working.
- Understand the shape of it. Read How It Works. Ten minutes there and the two-server design, the containers, and the nightly loop all click into place.
- Connect a channel. Ready to wire up your own account? Follow Playbook 1 — Connect a channel. Bluesky is free and the fastest to prove it out.
- Add a client. Turning it into a service? The client lifecycle playbook walks you from a new tenant block to their first approved post, including their private approval link.
- Learn the kill switch. Before you let anything run auto-fire, know how to stop it. One command —
touch /opt/portal/PAUSE — freezes all publishing instantly; remove the file to resume. Full detail in the stop-everything playbook.
One thing to note. Always reach the hub over
https (
postiz.wholetech.com). The login cookie will not stick over plain http — the single most common "I can't log in" surprise.
The one promise
If you remember nothing else from this manual, remember this. Automated marketing is only worth having if you can trust what it says without checking every line. That trust comes from one rule, enforced by the machine and not by good intentions:
It never states a fact it cannot verify. Every template declares the fields it needs. If a fact is missing from the source data, the post is held — shown to you, flagged, and never published as a guess. Real numbers or nothing. That is the honesty guard, and it is the reason it is safe to hand this engine your brand, or your client's.
Everything else in this system — the self-driving loop, the flat cost, the multi-tenant isolation, the approval option — exists to make that promise easy to keep at scale. Read on, and every claim here is one you can verify for yourself.
The companion manual
This manual covers the self-driving publishing system — the engine that writes posts from real data and ships them to your social accounts. Its companion, the Operator's Manual for the Auto Marketing Engine, documents the marketing department side in depth: the Auditor, SEO Manager, Content Desk, and the other named roles that read a page and produce the recommendations. Together they describe the whole system, end to end.
Two manuals, one systemRead this one to run the posting machine; read the
Operator's Manual to understand the departments that feed it.