Threads AI

Three working modes

From a file

You upload a list of finished posts and the bot publishes them in order. If there are several accounts, each takes its own post and nothing repeats.

AI generation

You set a topic and a few examples of the style you want — from there the bot writes the posts itself. The examples can be generated too, if writing them feels like a chore.

Smart mode

The same thing, except the bot analyzes the performance of its own publications and decides for itself what to write about and when.

How smart mode works

The bot collects statistics on each of its posts and calculates a simple score: views, plus likes at double weight, plus replies at five times weight. The logic is that a comment is worth far more than a view.

It then takes the top twenty percent of posts by that score and does two things. First, it adds them to its own pool of examples, so the next texts get written with those in mind. Second, it looks at which hours those posts went out in and rebuilds the schedule around those hours.

A trap we already walked into

The first version analyzed the account’s entire feed. If one account runs two scenarios — lifestyle and sales, say — posts from one got pulled into the examples for the other, and the lifestyle scenario started writing ads. Now each scenario analyzes only its own publications, and only the ones published after it was launched.

Working with accounts

Connect through a Meta login or by pasting an existing token — whichever suits you.
Tokens renew themselves. They live for sixty days; the bot tracks the expiry and renews ahead of time without your involvement.
Its own proxy per account, with a working check before it’s saved.
A persistent device fingerprint. It’s derived from the account identifier, so it survives both a restart and a loss of settings.
Pauses between accounts. Three to eight minutes, so publications don’t go out in a volley from one address.

Comments

The bot replies to comments under its own posts while holding the whole thread in mind: it can tell its own remark from somebody else’s. There’s a pause between replies and a per-session ceiling, so it doesn’t turn into a machine-gun commenter.

How it’s used

Let’s take a concrete task: we’re promoting an auto repair shop. We need a Threads account that writes useful posts about cars, builds an audience and brings people in to book a service.

You connect the account

In the Telegram bot you press «Add account» and log in through Meta or paste an existing token. The bot pulls the profile name in by itself and sets a timer on the token’s lifetime — after that it renews without you.

Account card

@garage_pro_minsk
Token: active, 58 days   Proxy: 195.x.x.10:8000
Auto-renewal: on

You create a scenario

You write the topic in plain words: «auto repair shop in Minsk, honest repairs, breakdowns explained, no upselling». Then you need example posts — either paste your own, or press «Auto» and the bot writes eight of them itself.

Scenario topic

Topic: auto repair, honest work, breakdowns explained
Example posts: 8 of them (generated)
Instructions for replies: short, to the point, no selling

You set the schedule

You specify how many posts a day and at which hours. Or you switch smart mode on right away — then the bot picks the hours itself, based on what performs best.

Schedule

3 posts a day→09:30→14:00→19:30

The bot starts publishing

Each post is generated fresh from the topic and the examples rather than pulled from a ready-made list. A random delay is added before publishing so posts don’t go out on the exact minute.

Example of a published post

Customer rolled in with «check engine light, must be the engine».
Actual cause — an oxygen sensor costing pocket change.
 
Moral: diagnostics first, panic second.
The check light comes on for fifty different reasons,
and a rebuild is nowhere near the top of that list.

It replies to comments

Under its own posts the bot replies by itself, holding the whole thread in mind: it can tell its own remark from somebody else’s. The pause between replies and the per-session limit are both configurable.

Discussion thread

@user_23: and what if the check light is on and it's misfiring?
bot: then it's coils and spark plugs next.
A misfire plus the check light is almost always ignition.
Pull the code — it points straight at the cylinder.

After two weeks the analytics kick in

The bot collects the metrics on its own posts, calculates the score by the formula «views + likes at double weight + replies at five times weight», takes the top twenty percent and tunes the topics and posting hours to match them.

Smart mode report

42posts published

1 380average reach

10:00 / 20:00best hours

case teardownleading format

What you need to start

A Threads account linked to a Meta app.
Proxies, if you have more than one account.
A topic and a few example posts in the style you want — or hand the generation to the bot.
A server, so the bot runs around the clock. We help with the setup.

Want Threads AI?

Leave a contact — we’ll show the bot working and help with the first launch.

We’ll reply within a day. No spam, no mailing lists.