Lite.panteo.ai logo
Lite.panteo.aiAgent Management System
super agent
Published: July 12, 2026

Super Agent: a full tour of the new-generation AI agent

Its own workspace, memory across conversations, file handling, GitHub, secrets and rollback. A complete walkthrough of what Super Agent — the AI agent inside Telegram and Max — can do.

Super Agent: a full tour of the new-generation AI agent

What Super Agent is

Super Agent is a new-generation AI agent that works right inside your messenger: in Telegram and in Max. There is nothing to configure, connect or train, unlike the bots of the previous generation. You write the task in plain words — it gets it done.

The key difference: the agent does not just reply with text, it does the work. It has its own workspace, files, memory and tools — which is why you can hand it long multi-step tasks, not just questions.

Its own workspace

Every user gets an isolated working environment where the agent actually does the work: creates and edits files, runs tools, checks the result. This is not a sandbox for an answer — it is a real environment where a task is carried to completion.

Isolation also works inside your own account: every chat, group and topic is a separate context with its own working folder. Your work project and your personal tasks stay apart.

Memory across conversations

The agent remembers context, projects and agreements. Come back to a conversation a week later and continue from where you stopped — no need to re-explain what happened.

If you want a clean slate, the /new command restarts the context of the current chat.

Working with files

Files are a normal part of the conversation:

  • You send documents, spreadsheets and images — the agent receives them into its working folder.
  • You get results back: anything the agent puts into its outbox arrives in your chat as a file.

So a task like "read this report and build a summary table" is solved entirely in the chat — no clouds, no folders, no forwarding.

Tools for real work

Command What it gives you
/github Connect GitHub through a one-time authorization link
/secrets Project keys and tokens via a one-time page instead of the chat
/rollback Roll project files back to the snapshot before your last message
/model Switch the model per task — per context, and it is remembered
/status What the agent is doing right now
/stop Interrupt the current run
/new Clean context
/credits Credit balance and trial status

Three of these are things chatbots simply do not have.

Rollback. The agent works with real files, so it needs an undo button. /rollback restores the project files to the state before your last message.

Secrets. Keys and tokens must not be pasted into a chat — they stay there forever. The /secrets command opens a one-time page that delivers credentials into the project workspace, bypassing the conversation.

Model choice. A fast, economical model for simple work; the strongest one for hard problems. Switch with /model, separately per context — the choice is remembered.

Working in groups

You can invite the agent into a work group. There it replies when addressed by name, and the owner decides who may assign tasks:

  • /invite <id> — allow a member to talk to the agent;
  • /uninvite <id> — revoke it;
  • /members — list who is invited.

A queue instead of interruptions

While the agent is working, new messages do not cut it off: they queue up and are picked up once the current answer is finished. Need to stop right now — /stop.

What it costs

  • Trial: 300 credits for 30 days — no card required.
  • Card linked: 8,000 credits for 30 days — effectively a free month of work.
  • After that: plans from 2,000 ₽ per month, paid in rubles, cancel any time with /unsubscribe.

You are billed for actual work: a fixed cost per active day plus pay-as-you-go for the model; days with no requests are free. The full breakdown is in «What an AI agent costs».

Read next

Frequently asked questions

What is Super Agent?
It is a new-generation AI agent that lives inside Telegram and Max. Unlike a chatbot it does not pick a scripted reply — it does the work in its own isolated workspace: handles files, remembers context across conversations, and carries multi-step tasks through to a result.
Do I need to configure or train it?
No. There are no scripts, no knowledge base and no integrations to set up — you simply write the task in plain words, the way you would to a colleague.
Does the agent remember previous conversations?
Yes. Every chat, group and topic is a separate context with its own working folder. Come back a week later and you continue where you left off, while work and personal tasks never mix.
Can I undo what the agent did?
Yes — the /rollback command restores the project files to the snapshot taken before your last message.