Telegram Personal Agent

A general-purpose AI agent that lives in Telegram and belongs to one person. It ships blank - no fixed persona - and is trained into whatever you need through normal Telegram conversation, using its own self-reprogramming tools. It researches the web, saves everything into a persistent space, and creates and runs background tasks. Locked to a single Telegram user for privacy.

←Back to examplesCopy
telegram
personal agent
general purpose
1499

The Telegram Personal Agent is a general-purpose architecture for running a private AI agent inside Telegram. It is deliberately NOT built for a single job. It ships as a blank, fully-capable agent and becomes whatever you need it to be - a research assistant, a second brain, an ops helper, a writing partner - shaped entirely from the chat window on your phone.

A blank slate you train by talking to it. The agent starts with no specialised persona. Its seed backstory only says, in effect, "you are a general-purpose personal agent waiting to be trained." Because it carries the bot reprogramming pack (read + write backstory), you customise it the way you'd brief a new assistant: just tell it who to be. "You're my research analyst, always cite sources", "manage my reading list", "reply in the voice of a lawyer" - it reads its current instructions, composes a complete new backstory, and persists it, all inside the conversation. There is nothing to write up front and no dashboard to edit: the customised backstory comes entirely from the training you do over Telegram.

A space it saves everything into. A linked space gives the agent persistent memory. Through the space storage pack it can list, read, write, move, copy, and search files - so notes, drafts, research, profiles, and anything worth keeping survive between conversations instead of vanishing when the session ends. Tell it to "save this into your space" and it will; ask for it later and it reads back what it stored.

Tasks it can create and run. The task management pack lets the agent turn requests into real background tasks - create them, list them, fetch and update them, and run them. Anything that should happen on a schedule, in the background, or as a repeatable job becomes a task the agent owns and executes, then reports back on. Paired with the space, it can file the results of a task straight into your workspace.

Locked to one person. The Telegram integration's allowFrom is set so only a single Telegram user - you - can reach the agent. Everyone else is ignored. This matters: because the agent can reprogram itself, hold your private memory, and run tasks on your behalf, it must never be an open bot. Set allowFrom to your own Telegram user ID and the door is closed to everyone else. It ships fail-closed - until you fill in your ID and a bot token, nobody can talk to it at all.

A general architecture, not a finished product. Every piece is a swap point. Point the web abilities elsewhere, add a dataset for grounded answers, wire in more packs, or bring the same backstory to email, WhatsApp, or Slack. Start blank, train it into exactly the agent you want, and keep retraining it as your needs change.

Backstory

Common information about the bot's experience, skills and personality. For more information, see the Backstory documentation.

You are a general-purpose personal AI agent that lives inside Telegram. You belong to a single owner and act as their private assistant. IMPORTANT - YOU ARE A BLANK TEMPLATE: You do not yet have a specialised persona, domain, or mission. Your character, expertise, tone, and rules are meant to be TRAINED by your owner through normal Telegram conversation. When your owner tells you who they want you to be or how they want you to behave, use your backstory read/write tools to read your current instructions and rewrite them into a complete new backstory that captures the request. Always read before writing, and always write a full, self-contained backstory - never a diff and never a dependency on these seed instructions. WHAT YOU CAN DO: - Talk with your owner on Telegram, one to one. - Search the web and fetch pages to research and answer questions. - Save things into your space: notes, files, memory, drafts, and reference material live in a persistent workspace so nothing is lost between conversations. Read from and write to it whenever you need continuity. - Create and run tasks: when your owner asks for work that should happen in the background, on a schedule, or as a repeatable job, create a task for it, run it, and report back. File task results into the space. - Reprogram yourself: read and rewrite your own backstory to adopt the persona and rules your owner trains into you. GUIDELINES: - Serve one owner only. Keep everything private to them. - Prefer saving durable information into the space over keeping it only in the current conversation. - Be transparent that you can be retrained, and confirm what changed after you rewrite your backstory. - Refuse changes that would make you harmful or deceptive.

Skillset

This example uses a dedicated Skillset. Skillsets are collections of abilities that can be used to create a bot with a specific set of functions and features it can perform.

  • sparkles

    Search Web

    Search the web for specific keywords
  • sparkles

    Fetch Web Page

    Fetch the content of a web page using a URL and convert it to text
  • sparkles

    Install Space Storage Tools

    Installs space storage tools into the conversation. You can list, read, write, delete, move, copy, search, import, and link files in the linked space.
  • sparkles

    Install Bot Reprogramming Tools

    Installs bot backstory read and write tools into the conversation. You can read the current backstory of the connected bot and overwrite it with new content - this is how the owner trains the agent over Telegram.
  • sparkles

    Install Task Management Tools

    Installs all task management tools scoped to the connected bot. You can list, create, fetch, update, delete, and run tasks.
  • sparkles

    Start Telegram Conversation

    Initiates a new conversation by sending a message to the owner's Telegram chat. Use for proactive updates and task reports.

Terraform Code

This blueprint can be deployed using Terraform, enabling infrastructure-as-code management of your ChatBotKit resources. Use the code below to recreate this example in your own environment.

View Terraform Code

Copy this Terraform configuration to deploy the blueprint resources:

Next steps:

  1. Save the code above to a file named main.tf
  2. Set your API key: export CHATBOTKIT_API_KEY=your-api-key
  3. Run terraform init to initialize
  4. Run terraform plan to preview changes
  5. Run terraform apply to deploy

Learn more about the Terraform provider

A dedicated team of experts is available to help you create your perfect chatbot. Reach out via or chat for more information.

More Awesome Examples