Updated: 2026-07-24
A bot with a neural network is a chatbot whose replies are generated by a language model rather than a pre-written script of buttons. It understands a question asked in your own words and answers by meaning. In Corpis such a bot is built with no code: you set the personality and rules in words, the neural-network brain composes the reply, and the same agent connects to Telegram by token.
Create an agent. Open agent creation and fill in the name and first message — or describe in words what the bot should be and the AI assistant fills the form for you.
Set personality and rules. In the agent settings describe the personality (who it is and how it speaks) and the rules of behavior. This is what drives the neural network's replies instead of a rigid script.
Check it in the preview. Message the agent in the preview and ask questions in your own words. Make sure the tone is right — adjust the personality and rules if needed.
Connect Telegram. Get a token from @BotFather with /newbot and paste it on the "Telegram" tab. The platform picks the message-receiving mode itself and shows a link to the bot.
Test the bot. Message the bot in a DM — it answers in your agent's voice. Add it to a group if you need to; there it replies when mentioned.
Behind the phrase is a simple thing: a language model (the "neural network") writes the replies instead of a pre-built script. Such a bot understands a freely worded question and answers by meaning rather than picking a canned line from a list.
In Corpis the neural network is the agent brain. It is provider-agnostic: the model is chosen in the settings and can be swapped without rewriting the agent, while personality, rules and body are stored separately from it.
Scripted builders assemble a bot from branches: a button — a pre-set reply — the next button. As long as the user follows the script it works; one step aside and the bot says "I don't understand" or sends them back to the start.
A bot with a neural network needs no script. You define not branches but personality and rules of behavior — the model builds the answer to a live question itself. Buttons do not disappear: you can add them as quick replies, but the dialog is not locked inside them.
The agent is created with a form: name, personality (who it is and how it speaks), rules of behavior and a first message. Writing the prompt by hand is optional — describe in words what the bot should be and the AI assistant fills the fields for you.
The finished agent answers right away in the preview, so you can check the tone and adjust the personality before connecting it to any channel.
The same agent becomes a Telegram bot by a token from @BotFather: get the token with /newbot, paste it on the "Telegram" tab — the platform keeps the connection itself, no server of your own is needed. In DMs the bot answers in your agent's voice; in a group it replies when mentioned.
You can start for free: there is a free plan and a 3-day Pro trial with no card required. Brain usage is counted in credits — each neural-network reply spends a share of them, and the limits are visible in the billing section. The free plan is enough to build an agent, test it and launch a Telegram bot; for a stream of messages you pick a paid plan.
It is a chatbot whose replies are written by a language model, not a pre-set script. It understands a freely worded question and answers by meaning, so talking to it feels like a conversation rather than picking menu items.
Corpis has a free plan and a 3-day Pro trial with no card required — enough to create an agent, test it and connect a Telegram bot. Brain usage is counted in credits, and the limits are visible in the billing section.
No. The agent is built with a form or a description in words, and Telegram connects by pasting a token. Code is only needed if you want the bot to call your own APIs — there is a separate "Commands and API" section for that.
A button bot works only along pre-written branches and gets lost on any off-script question. A bot with a neural network answers a free question by meaning; you can add buttons as quick replies, but the dialog is not limited to them.