Glossary
AI agent in Telegram: 18 terms in plain English

The words around AI agents come from three places at once: Telegram, language models and billing. Here they are in one list, each in two or three sentences, with the command you would actually type where there is one.

- AI agent
A program driven by a language model that takes a task in plain words and carries it out with tools: reads files, searches the web, writes code, sets reminders. A chatbot answers; an agent also acts.
- Context window
How much text a model can hold in view at once: your messages, files and its own notes.
/contextshows what fills it;/compactsqueezes the history and keeps the gist.- Default model
DeepSeek, included in every plan, Free too, with no keys to bring. Requests go through DeepSeek’s infrastructure in the PRC; you agree to that at onboarding.
- Forum topics
A Telegram group split into topics. With the agent added, each topic becomes its own session with its own memory, so several projects run side by side without mixing.
- Guest access
Letting another person use your agent in one topic:
/allow @username,/revoke @username,/allowedto list them.- Isolated container
The separate environment each agent runs in. Its files and memory sit there, and the agent knows only its owner and the people the owner lets in.
- Magnet ()
One unit of the agent’s work. Transcription is 1 per minute of voice, a spoken reply is from 4 per 1,000 characters, a web search 1.5 , a generated image 20 . Free gives 200 in the first month, then 100 a month; Pro gives 400 .
- MCP
Model Context Protocol: a standard way to plug tools into a model. The agent calls MCP tools the same way it reads a file or runs a search.
- Memory
Facts the agent keeps about you across chats: projects, people, preferences. It lives in your container, and you edit it in plain words: “remember”, “update”, “forget”.
- Mention mode
How the agent behaves in a group.
/mention on: it answers only when mentioned or replied to./mention off: it answers every message from its owner.- Onboarding bot
@wamagentsbot is where you start, pay and cancel. It creates your personal agent; after that you talk to the agent directly. Commands there:
/start,/status(subscription and the auto-renewal switch),/balance,/referral,/mybot,/support,/delete.- Own model
On Pro you can attach Claude with
/claudeor Codex on your ChatGPT subscription with/codex, signed in through the provider (OAuth). The provider then bills those tokens to your account./modelswitches models.- Reasoning effort
How long the model thinks before it answers:
/effortfromlowtomax, plusultraon Codex; for Claude and Codex. Higher is slower and suits harder tasks.- Scheduled task
A job the agent sets for later: a reminder for you, or a check it runs on its own at a set time. You ask in plain words: “remind me on Friday at 9”.
- Session
One running conversation with its context.
/resetstarts a fresh session with a new id;/clearclears the context, so the next message starts fresh in the same session.- Telegram bot
An account in Telegram run by a program instead of a person. Your agent is a bot of its own, with the name and avatar you choose.
- Transcription
Turning speech into text. Voice notes, audio files and the soundtrack of a video are transcribed automatically (Deepgram Nova-3).
- Vision
The agent’s ability to read photos and screenshots and answer about them. It runs through a separate service and costs 1.5 per image.
A
C
D
F
G
I
M
O
R
S
T
V
For the first minute with an agent, the guide walks through every command. What the agent does with all of this is in AI agent in Telegram.
Try it with your own agent
Free, no card. The agent starts working in the chat straight away.
Create your AI agentQuestions
What is the difference between an AI agent and a chatbot?
A chatbot answers with text. An AI agent takes a task and carries it out with tools: it reads files, searches the web, writes and runs code, sets reminders, and sends back the result.
Is a Telegram AI agent the same as a Telegram bot?
Technically it is a bot: an account run by a program. What makes it an agent is the language model and the tools behind it.


