What a trading bot actually costs, and what drives the number
The same sentence — “a Bybit bot” — covers a week of work and six months of it. Here is what moves the estimate, so you can shape the scope instead of receiving a number.
Vizanix engineering · about the author
- SECTION
- Why and when
- PUBLISHED
- 2026-08-30
- CHAPTERS
- 5
- READ NEXT
- 3
- LANGUAGE
- written in English
Every quote is a function of scope, and clients rarely know which parts of their description are the expensive ones. Usually the expensive parts are not the ones they emphasised.
What barely moves the number
- Which indicators. EMA, RSI, MACD, ATR — all cheap. Indicator maths is a rounding error in the estimate.
- Number of trading pairs. Once the universe is data-driven, one symbol and seven hundred cost nearly the same.
- Long or short. Both directions is the default, not an extra.
- Which venue, among the major ones. Bybit, Binance, OKX all have documented APIs. The first venue costs; the second is an adapter.
What moves it a lot
| Driver | Cheap version | Expensive version |
|---|---|---|
| Strategy definition | Written, precise, unambiguous | An idea to be formalised together |
| Order logic | Market in, stop and target out | Ladders, partial fills, multi-leg exits, hedge mode |
| Risk | Fixed size, one stop | Portfolio limits, correlation, drawdown, kill switch |
| Interfaces | Telegram commands | Web dashboard with roles and audit log |
| Accounts | One key | Multi-account, isolated parameters and statistics per key |
| Research | Client supplies validated rules | Backtest harness, data pipeline, model training |
| Data | Public candles | Recorded L2 book at event resolution |
The pattern: the trading idea is cheap; the operational surface around it is not. Most of the budget in a serious system goes into the parts that make it safe to run unattended, not into the parts that decide what to buy.
Why the first estimate is a range
An honest estimate before scoping is a range, and the range is wide because the biggest uncertainty is how precisely the strategy is defined. A one-sentence description can turn into two days of implementation or three weeks of joint formalisation, and nobody knows which until the questions get asked.
This is why our first milestone is deliberately small: a written specification and a prototype. After that the remaining estimate is a real number rather than a range, and you can stop there if the number is wrong for you.
The costs after delivery
Frequently omitted from the comparison and entirely predictable:
- Infrastructure. A VPS. Small money, but include it.
- Exchange API changes. Venues deprecate endpoints and change parameters. Something will need a fix each year.
- Strategy adjustment. You will want changes. Parameters are configuration; logic changes are work.
- Monitoring. Your attention, or a paid arrangement for someone else's.
- Data, if the strategy needs it. Recording L2 book across a universe adds up in storage.
How to get a cheaper bot without getting a worse one
- Write the strategy down precisely first. This is the single largest lever, and it costs you time rather than money. See how to write a spec.
- Cut the dashboard from phase one. Telegram control covers the first months at a fraction of the cost.
- Start with one venue. Multi-exchange is a real requirement or it is not; if it is not yet, defer it.
- Bring your own validated backtest. If you have already proven the edge, we are not paid to prove it again.
- Accept a narrower first version. One strategy, one market, running well, beats a platform that is nearly finished.
The inverse also holds: the most expensive projects we have seen are the ones that tried to build the general platform before knowing which specific thing it needed to do.
This article describes engineering practice. It is not investment advice. Vizanix develops software and does not promise trading returns.