Automation

From one thing you made to a channel that runs itself

Automation on Tonta is not a switch you flip on day one. It is what a workspace earns after it has shown Tonta what “good” looks like — and the rules for when a piece is allowed to post without you looking at it first are written down, checked every cycle, and never a black box.

How it starts

You don't configure a channel. You make one thing.

There is no service picker and no wizard on day one. You describe what you want in one box, the agent works out what kind of piece that is and fills in every setting it can, and you only get asked about the handful of things that actually change the output or the price.

Step 1 — You make one thing
Create mode: one box, no service picker.
A fictional creator typing one thing into a single box on a laptop at home, relaxed on a sofa.

Every default is real, visible, and open to change — never a blank field you have to fill in before anything happens. You get a priced plan before anything spends, and the piece it makes runs the same pipeline and the same approval gates as a fully configured channel. It is not a smaller or cheaper version of the real thing.

Step 2 — The offer, after three
Made with evidence, not on faith.
A fictional creator looking at her phone with a pinboard calendar visible behind her, pleased expression.

“You’ve made three of these. Want them to run every Tuesday and Friday?”

That is the actual offer, verbatim, after your third approved piece of the same kind — never before you have shown, three times, what you want. You accept with evidence, not on faith.

Saying yes is not a wizard. Every piece Create makes records the exact settings it resolved to, so graduating asks for only the difference — the rhythm and the platforms — and reuses everything else. Accepting means “keep doing that,” not starting over.

Once it's running

How often, and exactly when

A cadence never improvises its own schedule. Three things are decided upfront, in writing, before a single unattended piece goes out — so a missed slot is a choice you made in advance, not a surprise.

Content regularity
How many pieces per period, and whether they land as a batch up front or drip out one at a time.
Posting regularity
Per platform: which days, what times, in your timezone, and the minimum spacing between posts. The calendar shows the next fourteen days of slots, so an empty slot is visible before it is missed — never discovered after.
Blackout dates & missed slots
Holidays, launches, personal dates — nothing publishes in a window you block out. And what happens if a slot is still missed — skip it, or publish late — is a decision you make once, in advance, never improvised at 3am because a job finished behind schedule.

What still needs you

Three honest levels of trust

Unattended posting to your real account is a different kind of exposure than a more expensive feature — not a paid unlock, a risk you opt into once you've decided you trust the output. Every workspace starts at the safer end by default.

Review every piece
approve_each
Nothing publishes without you looking at it first. Every piece lands in review, every time. The safest option, and always available regardless of plan or history.
Review the first fiveDefault
approve_first_n
Every workspace starts here. The first five pieces on a new cadence go to review; once you have approved them, later pieces don’t need to wait on you by default. A new account earns trust before it publishes unattended — it is never assumed on day one.
Publish unattended
auto — a risk gate, not a perk
Available once your plan entitles unattended publishing and you have chosen to turn it on. This is not a feature you unlock by paying more — it is a different exposure you accept once you have seen enough of your own output to trust it. You can turn it off again at any time, for any piece.

Before every cycle

What has to be true before anything runs unattended

This is checked fresh, every cycle — never a stored yes. A connection can die, a budget can run out, a goal can already be achieved, and a stored flag would still say ready. The dashboard shows any blocker with the one action that clears it.

A goal, active

Unattended work with no target is just spending.

Required knowledge present

Whatever this service can't honestly run without — a service area, a glossary, a character sheet.

Complete cadence settings

How many pieces, how often, and where — nothing left to guess.

A healthy connection per platform

Including that none of them need reconnecting.

Required references present

The four services that need one to produce a professional result — not the ones where a reference is only a nice-to-have.

A budget with a cap

No cadence runs against an unlimited number.

An approval mode, explicitly chosen

Never assumed — you picked one.

Plan entitlement for unattended publishing

The risk gate below, cleared.

Money, on a schedule

A budget that warns you before it stops you

A cap that only speaks once it hits zero has told you nothing you could act on. Tonta forecasts continuously, so it can tell you a date rather than a percentage.

Forecast & warn
At current settings and current costs, Tonta forecasts when your budget runs out and warns you at 80% of the cap or at forecast exhaustion before the period ends — whichever comes first.
Pause cleanly at 100%
Scheduled runs stop and the cadence pauses. Work already in flight finishes and settles — it is already paid for, and nothing gets cancelled mid-flight to save money that is already spent.
Never a silent downgrade
Quality is never quietly reduced to squeeze inside a budget. If the numbers are tight, you get a choice instead — for example, switching to shorter pieces so the same budget lasts the month.

What it's aiming at

A goal that fits the channel, and a feedback loop you can see

Automation isn't just repetition — it is aimed at something, and what it's aimed at depends on what you're building.

A brand channel is measured on leads and sales. An episodic series is measured on completion and how much of the audience stays to the end. A reel channel is measured on throughput and cost per piece published. The goal picker offers what actually applies to what you’re running — never a generic list of vanity numbers.

And every cycle, the plan gets a little better on its own: whichever hooks, pillars, formats and lengths actually performed get weighted into the next plan — and you’re shown why, in plain language, not left to guess whether anything is learning at all.

Start with one piece. Graduate when you’re ready.

No wizard, no service picker, and nothing runs unattended until you’ve told it to.