🤖

Telegram and VK bots

Sales and support that never sleep

We build Telegram and VK bots for sales, support and accounting — with their own database and integration into the client's systems. At 3 a.m. the bot replies just as fast as at noon.

These are not business-card bots but production with transactions, migrations and live users: billing with payment confirmation, ticketing systems, document processing, accounting. Every bot comes with its own admin panel, logs and backups.

Problems clients bring

Typical situations and what we do about them.

Requests and payments get lost in chats

We move request intake, approval and records into a bot with a database. In a ticketing system a request moves through statuses and a resolution deadline; in subscription sales the payment receipt goes to the admins with 'Confirm' and 'Reject' buttons.

Documents and records are processed by hand

We build a bot that only needs a file forwarded to it: a PDF policy, a bank statement, a voice message or a photo of a receipt. It returns a finished file or text, and the result stays in the database.

You need a loyalty card, a queue or visit tracking

We replace paper cards and sign-up lists: a QR card in Telegram with a reward for visits, an electronic queue with reminders, and a list export to Excel.

Who it is for

Types of clients we have already built this for.

What we build

Types of work in this area. Each links to a case where it is already done.

Subscription sales bot with bank-transfer payment

The bot takes a customer from invite-code registration and a trial access to payment. It shows the bank-transfer details (Russian SBP system) and accepts a receipt screenshot, and the admins confirm the payment with a button. The subscription is then extended in the Marzban VPN panel, and the bot reminds the customer 3 days and 1 day before expiry. A person verifies the payment; this project has no automatic card acquiring.

Ticketing and service desk in Telegram

An employee files a request in four steps, and the card appears in the IT team's chat and changes status in place. It includes priorities and an SLA (a target time to resolve), closing only after the employee confirms, and a 1–10 rating. Alongside we build equipment tracking, an access registry with password encryption, a knowledge base and a monthly report for management.

PDF document processing by a bot

The bot accepts a PDF, analyses the page and returns a modified file. In our project it places an emergency commissioner's contacts onto an electronic motor-insurance policy while the policy number, personal data and QR code stay untouched, and a separate command checks the result. The document is processed in memory and never saved to disk.

Loyalty programme and electronic queues

A customer shows a QR code from the bot, an administrator adds a visit with one tap, and on the sixth night the bot unlocks a reward. A pause between accruals and a staff action log protect against abuse. The same approach works for an electronic queue: sign-up with one button, check-in, a notice when the turn is near, and a list export to Excel.

AI bot: speech transcription, statement and receipt parsing

Voice messages, receipt photos and complex text go to the multimodal Google Gemini model, which works with audio and images directly. This is how a voice-to-text bot with three modes and a finance assistant work: the assistant parses PDF statements from four banks and takes expenses as a plain sentence. Gemini keys run as a pool: when one key's quota runs out, the request goes through the next.

Media download and processing bot

A user sends a link and the bot returns the file in the same chat: yt-dlp fetches the video, FFmpeg compresses and splits it. Our project covers 12 platforms, quality choice up to 4K, MP3 with cover art, repeat requests from a cache and files up to 2 GB through a local Bot API. A separate bot downloads Hi-Res music and checks by spectrum that the FLAC is genuine.

Account sign-in and management bot for a service

The bot links a service account to a person in Telegram and issues a one-time code for passwordless sign-in. The owner approves registration requests with buttons, and changing a name or password is done from the chat. The bot has no database of its own: it calls the service's internal API, so the data lives in one place.

What you get

What stays with the client after handover.

  • A Telegram bot with button menus, roles and access rights
  • A database suited to the task: PostgreSQL, MariaDB or SQLite, tables created on start
  • An admin panel inside the bot: statistics, users, broadcasts, logs
  • Integrations with external systems over API: a panel, a service, an AI model, cloud storage
  • Docker Compose deployment on your server or ours; the bot version is written to the log on start
  • Database backups where the database holds the client's data
  • Documentation: a README written as a technical specification and a CHANGELOG
  • One month of free support after delivery, then maintenance by agreement

How the work goes

From the first request to handover: at every step it is clear what happens next.

  1. 1

    Request and free audit

    We go through the task: who uses the bot, which actions it must perform and where the data comes from. You get an estimate based on the scope of work.

  2. 2

    Scenarios and roles

    We describe the bot's dialogues and the rights of each role: what a customer, an employee, an administrator and a manager see. You get a description against which the finished bot can be checked.

  3. 3

    Database and integrations

    We design how the data is stored and connect external systems over API. You know where your data will live and what the bot takes from other services.

  4. 4

    Bot and admin panel development

    We build the bot, a button-based admin panel, broadcasts and logs. You try the working scenarios before the launch on the production server.

  5. 5

    Launch in Docker

    We deploy the bot on your server or ours and set up the time zone, logs and backups. You get a running service and a README with instructions.

  6. 6

    Delivery and support

    We hand over the project with documentation and a CHANGELOG and fix any issues free of charge for a month. After that, maintenance by agreement.

Stack for this area

Python and aiogram 3 →
An asynchronous bot answers many people at once while other requests are still being processed.
PostgreSQL, MariaDB, SQLite →
A database sized to the task: a ticketing system and record-keeping need a separate database server, while a small club is fine with a single SQLite file.
APScheduler →
A scheduler sends reminders, escalations and digests and makes backups on a timetable.
Google Gemini →
A neural network transcribes voice messages, reads receipt photos and parses complex text without separate recognition programs.
FFmpeg →
It checks a recording's length, compresses large files and splits video into parts so they fit Telegram's limit.
Docker →
The bot and its database start with one command on any server and do not depend on the machine's settings.

Pricing

Ballpark figures to understand the order of magnitude. Exact estimates come after the audit.

Simple bot
from $80
per project

Broadcasts, FAQ answers, bookings and request intake.

  • ✓Loyalty program and digital queues — from $160
  • ✓PDF processing, media download, AI bot via API — from $250
  • ✓Backup channel in VK
Support
from $100
per month

Monitoring, backups, small fixes and library updates.

  • ✓First month after delivery is free
  • ✓Feature requests are quoted separately
  • ✓Bot activity report on request

The audit and the proposal are free. No hourly billing: the price is fixed for the result, the exact figure comes in the proposal after the audit. Support after delivery starts at $100 a month, the first month is free. Timelines are individual and set once the specification is approved.

Cases in this area

12 projects — with the task, the solution and the numbers

In production · in-house product

Finance Assistant: PDF bank statements to a Telegram budget

our own product: a bot and web dashboard, four banks' statements, Gemini

In productionNDA

Telegram ticketing system for an IT department

for a museum complex in Moscow: tickets, asset tracking, knowledge base and analytics

In production · in-house product

Tech Poly VPN: VPN subscription billing in Telegram

our own product: SBP payments, key issuing in Marzban, a backup VKontakte bot

In productionNDA

Branding e-policies (OSAGO) with a Telegram bot

for accident commissioners: contacts on the form, data and QR code untouched

In production · in-house product

Golos Bot: Telegram voice messages transcribed to text

in-house Telegram bot: voice notes, video notes, audio and video to text via Gemini

In production · in-house product

ToshaMusic: Hi-Res music downloads with quality verification

in-house Telegram bot: FLAC 24-bit / 192 kHz from Qobuz, Deezer and Tidal

CompletedNDA

Loyalty card for a gaming club in Telegram

a QR code instead of a paper card: visit tracking and a reward for the sixth night

In production · in-house product

Streaming service access bot: code login from Telegram

our own bot for Peregovorka (ToshaStream): code login, approvals, roles

In production · in-house product

Downloader Bot: save TikTok and YouTube videos via Telegram

in-house bot: no watermarks, quality up to 4K, MP3, files up to 2 GB

CompletedNDA

Electronic lab-submission queue in Telegram

QueueBot for a university: one-tap sign-up, reminders and an Excel export

Client testingNDA

Telegram bot for selling VPN subscriptions

for a private VPN service: invite code, trial period, admin-approved payment receipt

In production · in-house product

Birthday bot: reminders and a gift budget

our own product: friends and family add the dates, the admin gets the reminders

Frequently asked questions

How much does a Telegram or VK bot cost?
A simple bot — broadcasts, FAQ answers, bookings and request intake — starts at $80. A loyalty program or digital queues start at $160. PDF processing, media download and an AI bot via API start at $250. A bot with payments and account management starts at $400, a ticket system with roles and SLA at $800. Support starts at $100 a month, the first month is free. We quote the exact price after a short free review of the task; there is no hourly billing.
Can you build a bot that tracks finances from bank statements?
Yes, we already run one: it accepts PDF bank statements, detects the bank and account itself, never double-counts overlapping periods and separates transfers between your own accounts from income and expenses. See the “Finance Assistant” case.
Will the bot run on your server or ours?
Whichever suits you. Usually we deploy on the client's server — code and data stay with you. We can also take it under our maintenance with monitoring and backups.
What if Telegram gets blocked?
For critical scenarios we build a backup channel on VK: the same bot, the same data. That's how access delivery works in our VPN service — if Telegram is unreachable, the config arrives via VK.
Can the bot be connected to our system over API?
Yes, a bot talks to other systems over API, that is, through a set of addresses by which one program asks another to perform an action. In our projects this is how the Marzban panel (creating users and extending subscriptions), a chat service's internal API and Google Drive for backups are connected. What to connect to your system, we work out during the free audit.
How can a bot take payments if there is no card acquiring?
The bot shows bank-transfer details, the customer sends a receipt screenshot, and the admins receive it with 'Confirm' and 'Reject' buttons. A safeguard stops two admins from confirming the same payment twice. If you need an automatic payment system, we discuss connecting it separately during the audit.
How do I build an AI bot that transcribes voice messages?
The audio goes to the Google Gemini model together with an instruction in a single request, with no separate speech-recognition step. Before sending, FFmpeg checks the recording and recompresses a heavy file. The audio is deleted right after processing, and if there is no speech in the recording, the bot says so outright.
How can employee and manager rights be separated in a bot?
The role is stored in the database and injected into every request, so a person never sees another role's menu. The ticketing system has four roles: employee, administrator, manager and owner. The manager gets reports and a dashboard but does not see the requests themselves.
Who owns the bot's code and data?
The code and data stay with the client. We deliver the project with documentation: a README written as a technical specification and a CHANGELOG. After delivery there is a month of free support, then maintenance by agreement.
How long does it take to build a bot?
It depends on the scope: the number of roles, integrations and scenarios. We name the timeline after a short free audit, once it is clear what exactly has to be built.

Need something from “Telegram and VK bots”?

A short review is free: we'll look at the task and tell you what's worth doing and roughly for how much.