Web apps, bots and automation for business. We take your project from the first conversation to launch and support.
We are Grahovskiy: development, integrations and marketing in one team. We build websites, online shops and CRM systems, connect them to payment and accounting services, automate routine work and bring customers through ads and SEO. Over 7+ years we have built the system and website for a clinic in Barcelona, a marketplace in Belarus, Telegram services and product-marking integrations. You talk directly to the people building your project, get a clear timeline and estimate before we start, and one month of free support after launch.
We take the whole job: from the website and CRM to the ads that bring customers.
Business-card sites, landing pages and small company websites. Fast launch, responsive layout, request forms, integration with CRM and messengers.
By quote
E-commerce platforms and marketplaces: catalog, seller accounts, checkout, payments and delivery.
By quote
We bring customers to what we build: targeted and search ads, SEO, social media and email campaigns, with analytics that shows where leads come from.
By quote
The system a company runs on: client records, services and estimates, the team, enquiries and channels.
By quote
We connect your systems with payment, labeling and accounting services: Chestny ZNAK, 1C, warehouses, marketplaces, SMS and email.
By quote
We automate routine work: documents, reports, bookings and registrations, syncing CRM, spreadsheets and messengers.
By quote
We collect data from websites and marketplaces: competitor prices, stock, product cards, auction lots. Data arrives on schedule — in a spreadsheet, your CRM or Telegram.
From $300
A bot that sells and serves customers on its own: takes requests and payments, issues access, sends notifications and passes data to your CRM.
From $250
Ongoing support and development of your product: monitoring, updates, new features.
By quote
Every stage ends with a result you can check.
The system a clinic runs on: patient records, a service price list and estimates, doctors and schedules, ad enquiries and WhatsApp. What was built: • A patient record: contacts, source and the history of work. An incoming enquiry becomes a lasting record. • A service price list and an estimate: the plan of work and the prices come from the clinic price list and go to the patient as a document. • Doctors, specialties and schedules. A staff member sees their own patients, a manager sees the whole clinic. • WhatsApp: a message lands on the record, the reply goes out from the system, and the delivery status stays with it. • Submissions from Facebook and Instagram forms create a record on their own. A doctor's calendar is linked to Google Calendar. Result: the clinic runs patients, services and the team in one system. Stack: Odoo 19, PostgreSQL, WhatsApp Cloud API, Meta Lead Ads, Google Calendar.
The clinic website in Barcelona: dentistry and aesthetic medicine on the same pages, a consultation booked in WhatsApp. What was built: • A home page and two direction pages. A visitor chooses where to go at once. • The booking button opens WhatsApp with the clinic's number. • Photos of the doctors, the clinic and the work, and video on the first screen. • The same screens on a phone as on a computer. Result: a patient books from the site. Stack: HTML, CSS.
A system for a client who analyses vehicle damage from US salvage auctions. The goal was to collect photos of thousands of lots and organise them into an archive that is easy to work with. What was built: • A Telegram bot: the client sends an auction export (Excel, CSV or zip), the bot selects the matching cars and downloads every photo of each lot. • Filtering by year, makes and damage types, adjustable with buttons right in the bot — no developer needed. • Tidy storage: Damage / Make / Year / Model / Lot_VIN, with a data file per lot — VIN, trim, damage, mileage, title, sale date. Source tables are kept unchanged, with checksums. • Live progress, stop and resume, no duplicates, a disk space check before every download. • Robust to real-world Excel files: multiple sheets, CSV from localised Excel, duplicates, extra columns. • The archive mounts as a network drive on Windows and macOS over secure SFTP. • The server is provisioned and updated with a single command: Docker, firewall, brute-force protection, key-only login. Runs on a 1 GB VPS. • 195 automated tests; a deployment stops automatically if any test fails. Result: from a daily export of 50,000 lots the bot picks the relevant ones and builds the archive — 12–13 photos per car on average, with no manual work. Stack: Python 3.12, aiogram 3, openpyxl, Docker, Ubuntu, OpenSSH/SFTP, pytest.
A marketplace where Belarusian sellers open their own stores and buyers order goods and services with nationwide delivery. We built the marketplace from scratch in 2024 and keep developing it — from order and payment architecture to tools for sellers. Live: crossvill.by. What was built: • Multi-vendor model: every seller has their own store, catalogue and orders. A cart with items from different sellers is automatically split into sub-orders. • Online payments via WebPay with split payouts: money is distributed between sellers, and marketplace and bank fees are calculated automatically. • Commissions per category and per seller, with validity periods; a referral programme for partners. • Delivery by Belpost and Evropochta, with pickup point selection through their APIs. • Bulk product upload from Excel: templates, creating and updating products with variants and attributes, prices and stock, SKU renaming. Processing runs in the background with a per-row error report. • Product photos and videos are pulled from Yandex Disk links; storage in S3, images served via imgproxy and a CDN. • Product ranking by clicks, purchases and freshness, recalculated nightly. • Financial reports by order and seller with stored calculations; new seller applications go straight to Google Sheets. • Background jobs on Solid Queue, error monitoring in Sentry, feature flags, deployment to Docker Swarm behind Traefik. Result: a live platform where sellers manage their own catalogues and get paid, while the marketplace automatically retains its commission. Stack: Ruby 3.3, Ruby on Rails 7.1, Spree Commerce, PostgreSQL, Hotwire, Solid Queue, S3, imgproxy, Docker, Traefik, Sentry.
A bot through which the client sells VPN subscriptions: customers pick a plan, pay and instantly receive a key — no administrator involved. Built from scratch, later with a separate version for VK. What was built: • Three protocols: WireGuard, Outline and VLESS. The key is delivered right after payment — as a file, a link and a QR code. • New servers are added from the admin panel: the bot connects over SSH, installs and configures the VPN itself, and alerts administrators immediately if something fails. • Choice of location and protocol; switching can be offered as a paid option. • Eight payment methods: YooKassa, YooMoney, Tinkoff, Telegram Stars, Cryptomus, WalletPay, FreeKassa and AAIO — with automatic payment checks and subscription auto-renewal. • Trial period, promo codes and a partner programme with payout tracking. • An admin panel right inside the bot: users, payments, charts, server speed tests, backups. • All bot texts live in a separate file and can be edited without a developer. • A separate version of the bot for VK built on vkbottle. Result: the service runs on its own — from payment to key delivery and renewal; the administrator only adds servers and watches the stats. Stack: Python, aiogram, pyTelegramBotAPI, vkbottle, SQLite, Paramiko (SSH), WireGuard, Outline, VLESS, YooKassa API, Plotly.
A service for a client who needed to find out about free appointment dates at Lithuanian consulates in time: slots appear rarely and are gone quickly. What was built: • Regular checks of the online appointment schedule across three consulates at once. • Telegram notifications as soon as a free date appears, with the consulate and the date. • A history of found slots: what appeared and when, with no repeat alerts for the same slot. • Detailed logs and page screenshots on errors, so problems are visible immediately. • Migration from Selenium to Playwright with an asynchronous architecture: more stable, faster and lighter on resources. Result: the client learns about a free slot within seconds of it appearing and has time to book it, instead of refreshing the page by hand. Stack: Python, Playwright (previously Selenium), asyncio, TinyDB, Telegram Bot API.
A library for a company selling goods subject to mandatory marking: it automates marking in Russia's Chestny ZNAK system — no more manual work in the web portal. What was built: • The full marking cycle via the SUZ (OMS) and True API (GIS MT) interfaces: ordering codes, retrieving them, the application report and putting goods into circulation. • Authorisation and electronic signature: tokens are obtained and documents are signed with a qualified electronic signature (CryptoPro) through a separate signing service. • Product (GTIN) list loading with automatic pagination. • Waiting for orders to be ready and reports to be processed, document status checks — the system waits for Chestny ZNAK's response on its own. • Correct handling of marking codes, including GS1 control characters. • Clear errors and timeouts at every step; a modular design, so each step can be called on its own or embedded into the client's business system. • Documentation and examples for the client's team. Result: a batch of goods is marked with a single call from the client's system instead of manual operations in the portal. Stack: Python, requests, SUZ API 3.0, True API (GIS MT), CryptoPro.
A permanent make-up studio in Minsk. Before us, posts were boosted with the "Promote" button: about 20 enquiries a month at ~$12. Images: sample creatives in the project's style. What we did: • Before/after Reels and a 15% first-visit offer. • Women aged 25–45 within 5 km, enquiries straight to direct messages. • Retargeting people who watched the Reels and profile but did not book. Result: 60–80 bookings a month at ~$5 instead of ~20 at $12. Budget ~$300–400 a month, about 6 months. Stack: Meta Ads, Instagram, pixel, retargeting.
A showroom for made-to-order kitchens and furniture. Before us: only organic reach and word of mouth, about 10 enquiries a month. Images: sample creatives in the project's style. What we did: • Carousels with real projects and prices. • A quiz form instead of a plain enquiry form. • Retargeting people who watched the videos to the end, plus running the brand pages. Result: 35–50 enquiries a month at $14–18, including 12–15 designer home visits. Budget ~$600–700 a month, about 8 months. Stack: Meta Ads, Instagram, SMM, retargeting.
A used-car dealer. Facebook only, the goal is enquiries for finding a car. Images: sample creatives in the project's style. What we did: • Carousels with specific cars and prices. • A "trade-in and a loan with no down payment" offer. • A filter question in the lead form to screen out irrelevant enquiries. Result: 120–150 enquiries a month at ~$5–7; about 7% turn into a sale, 8–10 cars a month. Budget ~$800 a month, about 4 months. Stack: Meta Ads, Facebook, lead forms.
The CrossVill marketplace with delivery across Belarus. We built it and run its ads on Google, Yandex and Meta. Images: sample creatives in the project's style. What we did: • Google Shopping as the main sales channel. • Product campaigns in Yandex Direct. • Meta for retargeting people who opened a product and did not buy. Result: 250–300 orders a month, ROAS ~6.5 (ad spend ~15% of revenue). Budget ~$1,500 a month, about 10 months. Stack: Google Ads, Google Shopping, Yandex Direct, Meta Ads.
An Amazon seller. External traffic from Google and Facebook to the product listings. Images: sample creatives in the project's style. What we did: • Google Search on "category + feature" queries. • Sales tracked through Amazon Attribution. • Video retargeting on Facebook. Result: 350–450 sales a month from external traffic, ROAS ~3 (ad spend ~33% of revenue). Budget ~$1,000 a month, about 5 months. Stack: Google Ads, Meta Ads, Amazon Attribution.
Depends on project complexity. Telegram bot - from 1 week, web scraper - from 3 days, web application - from 2 weeks. Exact timelines discussed after requirements analysis.
Usually 50% upfront, 50% after delivery. For large projects, stage-based payment is possible. We accept bank transfers and cryptocurrency.
Yes, first month of support is free. After that by agreement or subscription fee.
Tell us about your task — we will reply within a day with questions or a first estimate.