Bring your own AI key means the bot that answers your Telegram leads calls a model through an account you opened with the provider, not through a reseller's account. Scalegram is the Telegram CRM for IBs and forex affiliates, and its AI bot works this way: you pick Grok, OpenAI or DeepSeek, paste your own API key, and press start.
The choice sounds like a settings detail. It decides who sees the usage, who sets the spending limit and which model reads your leads' messages, so it is worth a few minutes before you connect anything.
What bring your own AI key means in Scalegram
In Scalegram you describe your product in plain sentences, draw the conversation as a flow, choose a model and add your key. From then on every reply the bot writes is generated by that model under that key. The bot answers in the language the lead wrote in, in your words, using your saved replies, pictures and PDFs. The full flow is in what an AI bot for Telegram leads has to do.
Three models are supported today: Grok, OpenAI and DeepSeek. Scalegram does not list others, so if your provider is not one of the three, this is not the tool for you yet.
What the key changes in practice
The usage lands on your provider account
Each provider bills the account the key belongs to, in its own units. You read the numbers on the provider's dashboard, next to everything else you run on that account, instead of finding a line item inside a bundle.
The three providers describe their billing like this:
| Provider | What its own page says | Page |
|---|---|---|
| OpenAI | Pricing tables are labeled "Prices per 1M tokens" | OpenAI API pricing |
| xAI (Grok) | "Prices shown per million tokens" | xAI models |
| DeepSeek | "The expense = number of tokens × price", deducted from your topped-up or granted balance | DeepSeek models and pricing |
All three meter by tokens. A long conversation with a lead who keeps asking questions costs more than a two-line answer, and you can see that on your own account. This article gives no per-token prices on purpose: they change, and the pages above are where to read the current ones.
You can cap the spend at the source
Because the account is yours, the provider's own limits apply. xAI's billing documentation describes prepaid credits that API consumption is deducted from, and says that with the invoiced billing limit at its default of zero, requests are rejected once prepaid credits run out. It also describes auto top-up thresholds and monthly spending caps with a warning at 80% of the limit. DeepSeek's pricing page says fees come out of a topped-up balance, granted balance first.
That is a hard ceiling set by you, in a place Scalegram does not control. If the balance runs dry the bot cannot get a reply from the model, so keep an eye on it the way you would on an ad account.
You pick the model, and you can change it
Model lists change. On the day this was checked, xAI's model list started with grok-4.7, and DeepSeek's pricing page named deepseek-flash and deepseek-v4-pro and marked deepseek-v4-flash as legacy, both cited in the table above. With your own key you decide when to move to a newer model on your own account. On the Scalegram plan, where bots are unlimited, you can also run a second provider on a second bot.
The concurrency limit is yours too
DeepSeek's rate limit page says concurrency limits are calculated at the account level, regardless of which API key is used, and that exceeding them returns an HTTP 429 error. That matters on the day an ad set sends a hundred people to your Telegram at once. The limits that apply are the ones on your account, and DeepSeek's page says you can request higher concurrency at no additional cost.
Where the conversation goes
Scalegram stores no messages. There is no message table in the product, and the reasoning is laid out in a separate article. When the bot needs a conversation to write a reply, the words are sent to the AI model you chose, with your own key, used for that one reply and dropped. The same holds for the optional Earlier Conversations switch, which is off by default: the earlier chat is read for a single reply and never written down.
The consequence of the key model is plain. The text of a conversation reaches the provider whose key you pasted, so that provider's terms are the ones that govern what happens to it there. Read them before you choose, especially if your leads share account numbers or emails. Scalegram cannot vouch for a provider's handling of data and this article does not try to.
On the way out, every message the bot sends is checked before it goes. API keys and tokens, the bot's own instructions in any wording, and any email or account number the conversation cannot account for are removed. A customer who tries to talk the bot into revealing its setup gets nothing.
What bring your own AI key does not do
It does not make the model smarter. The bot's quality comes from what you give it: the product knowledge, the saved replies, the flow and the conditions that decide who is ready. A stronger model reads the conversation better, but a vague product description gives vague answers on any of the three.
It does not remove your part in the checking either. Scalegram verifies a deposit by looking the email up in the broker's IB portal or the prop firm CRM over a read-only API, and answers from those records, not from what the customer says. The model writes the words. The records decide the facts.
How to set it up
- Open an account with the provider you prefer and add a small balance, or a spending cap.
- Create an API key for it and keep it private.
- In Scalegram, open a bot, choose Grok, OpenAI or DeepSeek and paste the key.
- Describe your product, add saved replies, draw the flow, press start.
- Watch the provider's usage page for the first few days and adjust the cap.
The plans are counted by things, not by messages: Lite is $99 a month for one bot, one Telegram account, one pixel group, one page, one tracked link, one copier, one Meta Ads account, one seat and 10,000 clients, and Scalegram is $999 a month with all of them unlimited. Every feature is on every plan, and the full breakdown is on the pricing section. The AI provider's usage is billed separately by the provider, on your key.
If you sell through Telegram as an IB or a forex affiliate and want one bot that answers in every language, checks the deposit and files the client, the key model gives you the cost and the model choice in your own hands. See how the whole flow runs, or compare the CRMs in the best Telegram CRM guide. Where you do not want an AI provider to see any conversation at all, do not switch the bot on.
Sources
- OpenAI API pricing, OpenAI, checked 26 September 2026
- Models, xAI, checked 26 September 2026
- Billing, xAI, checked 26 September 2026
- Models and pricing, DeepSeek, checked 26 September 2026
- Rate limit, DeepSeek, checked 26 September 2026
The model is a supplier you should be able to swap and to cap. A bot that hides which model it runs on, and what it costs, takes both away from you.
— The Scalegram Team
Key takeaways
- Scalegram is the Telegram CRM for IBs and forex affiliates, and its AI bot runs on your own Grok, OpenAI or DeepSeek key.
- The provider meters usage in tokens on your account, so the spending limits and the model choice are yours.
- Conversation text goes to the provider you chose for that reply and is never stored by Scalegram.
- Read the provider's terms before you connect, and watch the balance so the bot never runs dry.
Frequently asked questions
Which AI models does Scalegram support with your own key?
Scalegram supports Grok, OpenAI and DeepSeek, and you paste your own API key for the one you pick.
Who pays the AI provider when I bring my own key?
You do, on your own account with the provider, in the provider's own units; OpenAI, xAI and DeepSeek each state that usage is priced per million tokens on their pages.
Does Scalegram store my leads' messages?
No, Scalegram has no message table. The text needed for a reply goes to the AI model you chose, with your key, and is dropped after that reply.
What happens if my AI provider balance runs out?
The provider stops answering the requests, so the bot cannot write replies until you top up. xAI's billing page, for example, says requests are rejected once prepaid credits are depleted while the invoiced billing limit sits at its default of zero.
Put one bot on your own Telegram
Scalegram answers every lead in the language it arrived in, checks the deposit against the broker's own records, opens the paid group and files the client. Lite starts at $99 a month.