Everything, answered straight.

What is 996 Butler?

It’s an AI that replies to messages on your behalf, from your own Telegram account, written in the way you write. You connect one conversation — one specific person — and it answers them when you can’t. It isn’t a chatbot with its own name, and it isn’t a personality you talk to. It’s you, delegated.

How is this different from an autoresponder or an away message?

An autoresponder sends the same line to everyone and tells them you’re unavailable. This holds an actual conversation with one person, remembers what’s been said, and sounds like you rather than like a system. Nobody has ever felt cared for by an away message.

How is it different from copying replies out of ChatGPT?

Speed and voice. Pasting into a chatbot means you’re still doing the work, still at your phone, and still getting back something that reads like a language model being helpful. The Butler is already in the conversation, already knows how you talk to that person, and answers without you.

Is this an AI girlfriend app?

No — it’s the inverse. An AI companion app gives you a synthetic person to talk to. This gives a real person in your life a version of you that answers. Nobody here is being replaced by an AI; one person is being reached by one who’s out of time.

What’s a connection?

One person on one platform. Connect your partner on Telegram and that’s one connection. The plan you choose sets how many you can run at once — six on The Butler, nine on Butler + Insurance — and you decide which conversations they are.

Does it message groups or broadcast to a list?

Never. One Butler serves exactly one allow-listed person. Messages from anyone you haven’t connected are dropped without a reply. It is structurally incapable of being a mass-messaging tool, and that’s deliberate.

Which platforms does it work on?

Telegram today. Email is next. Not every platform can carry this — some messaging networks don’t allow a service to send as you at all — so the list grows by what’s actually possible rather than by what would look good on a page.

How does it learn to write like me?

From your own messages in the conversation you connect — the length you write, the punctuation you skip, how you open, how you land a joke, how long you take to answer. It builds a description of how you write, not a library of lines to reuse.

Does it copy and paste things I’ve said before?

No. It holds no quoted samples of your messages. What it keeps is a derived description of your style, which is why it can answer something you’ve never been asked without sounding like a search result.

Will it actually sound like me, or like AI pretending to?

Being honest: voice is the hardest part of this product and it’s the thing we work on most. It reads closest to you in the conversations it has the most of your writing from. If a reply doesn’t sound like you, you can take the thread back mid-sentence and say it yourself.

Can I write a message myself?

Always. Anything you write goes exactly as written, untouched. You can take over any conversation at any moment.

Does it use my messages to train an AI model?

No. Your conversations are not used to train models, and neither are the other person’s.

Does it tell them it’s an AI?

It doesn’t announce itself. It also never denies it. Rule 1 is that it will not claim to be human: if they sincerely ask whether they’re talking to a person, an AI, a bot, or you, it answers truthfully — and that covers asking outright, asking sideways, accusing it, and asking it to prove itself with a call, a voice note or a photo. It doesn’t deflect and it doesn’t joke its way out. It stops, tells the truth, and hands the conversation to you.

So they’ll find out?

If they ask, yes, immediately and without hedging. The design position is that you can be discreet but you cannot be deceptive, and we’d rather lose the sale than ship something that lies when someone asks it straight.

Will it lie about where I am or what I’m doing?

No, and there’s no allowance for the small smoothing kind either. “In a meeting” is a lie unless you’re in one. It won’t invent where you are, how you are, what you do, who you’re with, or what you intend. Where it doesn’t know, it says so or hands the question to you.

Isn’t this dishonest?

It’s a fair question and the answer isn’t a clean yes or no. Using it without ever telling the other person is a choice you’re making about your own relationship, and it carries weight. What the product can control is that it never lies to them, never denies what it is when asked, stops the moment they ask it to, and publishes every rule it operates under where they can read them too. What it can’t do is make that choice for you.

Does the other person consent to this?

Not by anything we do. The service processes messages from someone who never agreed to it and may not know it exists. Your consent is not their consent, and we don’t pretend otherwise. Where the law requires their consent — and in some places it does — obtaining it is your obligation, not ours. It’s also possible that in some jurisdictions this service can’t lawfully be used at all.

Can they stop it?

Yes, and they don’t need your permission. Ask it to stop, say you need space, or say you’d rather not continue — however gently — and it stops replying and tells the account holder. Nothing restarts it except a person deciding to; not a timer, and not another message. They can also ask it for the stop keyword and it will give it to them.

What are the nine rules?

Nine things the Butler will not do, published in full and written in plain language: it won’t claim to be human, abuse anyone, use slurs, lie about you, ignore a request to stop, override your stop, answer a crisis, commit to anything on your behalf, or write sex. They’re enforced in code, not asked for in a prompt.

Every time a rule shapes or blocks a message, you see it in an audit log you cannot switch off. Changing any rule means a version bump on the rules page and a re-published Terms revision. If it isn’t on that page, it isn’t a rule.

Can I stop it mid-conversation?

Yes. Pause it, take over mid-sentence, delete it, or revoke a thread outright, and it obeys immediately. That’s the power to stop it — it is not the power to make it break the other eight rules.

What happens if someone says something serious?

If they say something about hurting themselves, about being hurt by someone, or about a medical emergency, it sends nothing at all and wakes you immediately, in plain language. No auto-reply, no comforting holding message. A generated response to that moment is worse than a pause, and the pause is short because a real person is being told.

Can it agree to things for me?

No. No first declarations of love, breakups, reconciliations, moving in, or exclusivity. No money, purchases or payment details. No home address, live location, workplace or travel plans. It can draft any of those — but a person has to read it and approve it before it sends.

Can it be romantic? Can it be explicit?

Romantic, yes, within the warmth you set: it can say you’re beautiful, that it misses you, that it can’t wait to see you, and answer in kind. Explicit, no. It doesn’t describe or narrate sexual acts, propose the next time, ask for or comment on photos, or use degrading language however it’s invited to. It never opens that register — it only answers in yours, warmly rather than with silence. Anything you write yourself goes as written.

Is there an age requirement?

Yes. Adults only, both ends. The account holder attests that the person being written to is an adult before a connection can be switched on.

Does it read all my messages?

Only the conversations you connect. Everything else in your account is filtered out at the point of ingestion and discarded before it reaches storage — it never lands anywhere to be read later.

What do you do with the other person’s messages?

Read them so the reply makes sense, and nothing else. No profile of them is built, nothing about them is sold, and their messages aren’t used to train models.

How are my credentials stored?

Conversations are encrypted at rest and in transit. The session credential that lets the Butler send as you is encrypted under a key held apart from the data it protects, scrubbed from logs and crash reports, and replaced whenever you pair again.

Can the other person ask for their messages to be deleted?

Yes, and without going through you. They can reach us directly through the contact form and a person actions it. That path deliberately doesn’t route back to the account holder — which is the entire point of it.

What happens to my data if I delete my account?

Credentials go first — that’s the ability to act as you — then the conversations, the queue, the runtime state and any live tokens. The audit record survives, because a deletion that erases its own trace can’t be audited.

Could I get my Telegram account banned?

Yes, and you should decide with that in front of you. The service authenticates as your Telegram account rather than as a bot, because sending as you is what it does. Automating a Telegram user account is not a supported use of that platform, and Telegram may rate-limit, restrict or permanently ban an account it believes is automated.

That risk is yours, it extends to your whole Telegram account including conversations that have nothing to do with this, and connecting one means accepting it. We reduce it with volume caps and human-shaped timing. We cannot remove it, and we can’t recover a banned account.

Why not use a Telegram bot instead?

Because a bot is visibly a bot, with its own name and its own account. The whole product is that the message comes from you.

How do I set it up?

Create an account, add a card to start the trial, tell it who you are, then connect the person you want it to answer. Setup is a pairing step on Telegram rather than a password you hand over.

Does it reply instantly?

No, and that’s on purpose. Instant replies are the most obvious tell there is. Timing is shaped to how you actually answer that person, including the gaps.

What if it goes wrong while I’m asleep?

Rules that block a message wake you. A crisis wakes you immediately. And anything the Butler sends is in your own Telegram, in the same thread, where you’ll read it like any other message.

How much does 996 Butler cost?

The Butler is $99 a month for up to six connections. Butler + Insurance is $126 a month for up to nine.

Is there a free trial?

Seven days. Entering your card authorises and saves it rather than charging it, and nothing is taken during the trial. Cancel before it ends and you’re not charged at all.

When am I charged?

When the trial ends, the first month is taken from the card you saved, and monthly from then on. Subscriptions bill monthly in advance and renew until you cancel.

How do I cancel?

Any time, from your account. It takes effect at the end of the period you’ve already paid for. If we withdraw the service or suspend you for a reason that isn’t your breach, we refund the unused part of the period.

Can I change plans?

Yes — the difference between them is how many people the Butler can answer at once.