The first question most traders ask about a crypto bot is which strategy it runs. The question they should ask first is where it runs. Cloud-hosted bots and self-hosted bots make fundamentally different trade-offs around security, cost, control, and reliability. These trade-offs matter more than strategy selection for traders managing significant capital.
The Security Difference
Cloud-hosted platforms require your exchange API keys to be stored on their servers. When you connect Binance to a cloud bot platform, those credentials are transmitted over the internet and stored in the platform's database. The platform promises encryption and security best practices. You trust their implementation.
Self-hosted platforms store API keys on your own hardware. Our platform uses Fernet encryption with PBKDF2 key derivation for API key storage. The encrypted keys sit in a local SQLite database on your machine. They never travel over the internet. No third party has access.
The risk calculus is straightforward. A cloud platform is a centralized target. If breached, every user's exchange credentials are compromised simultaneously. Cloud bot platform breaches have occurred in the crypto space, exposing thousands of API keys. Even with trade-only permissions (no withdrawal capability), an attacker with your trade API keys can place destructive trades, liquidate your positions, or front-run your strategies.
A self-hosted platform is a distributed target. An attacker would need to compromise your specific machine to access your specific keys. The attack surface is limited to your personal security hygiene (strong passwords, updated OS, firewall configuration) rather than a centralized platform's entire infrastructure.
The Cost Difference
Cloud platforms charge subscription fees, typically 15 to 100 dollars per month depending on the tier, plus percentage-based fees on trading volume at higher tiers. For 45 bots trading actively across 13 symbols, monthly costs on a cloud platform would range from 50 to 200 dollars depending on volume and tier.
Self-hosted platforms run on hardware you already own or can provision cheaply. Our entire system, 45 bots, a FastAPI server, a React dashboard, SQLite database, and local Ollama AI model, runs on a single Mac. The marginal cost of running the trading system is the electricity it consumes, which is negligible. There are no subscription fees, no volume fees, and no per-bot charges.
The Claude API for AI enrichment and sentiment analysis costs a few dollars per month at our usage levels (under 200 calls per day with significant cache hit rates). This is optional and the system operates fully without it.
Over a year, the cost difference between a cloud platform at 50 dollars per month and a self-hosted platform at near-zero marginal cost is 600 dollars. For a 45,000 dollar portfolio, that is 1.3 percent of capital consumed by platform fees alone, directly reducing returns.
The Control Difference
Cloud platforms restrict what you can do. Strategy options are limited to what the platform offers. Risk management is limited to the platform's built-in features. Backtesting, when available, uses the platform's engine with its assumptions about fees, slippage, and market impact. You cannot inspect the source code, modify the risk logic, or verify the backtesting methodology.
Self-hosted platforms give you complete control. Our risk framework has five sequential per-bot checks, three portfolio-level constraints, automatic decay detection, and a dead man switch. We built each of these because we needed them. A cloud platform's risk management is whatever their product team decided to build, and you have no way to extend it.
Backtesting on our platform uses 0.10 percent taker fees on both sides, 2 to 10 basis points of slippage, and realistic position sizing against current equity. We can verify this by reading the source code. On a cloud platform, the backtesting engine is a black box. The fee assumptions, slippage model, and sizing logic are hidden. Backtest results might be optimistic (understating fees) or pessimistic (overstating slippage) and you have no way to know.
The Reliability Trade-Off
Cloud platforms have one genuine advantage: uptime management. They run on professionally managed infrastructure with redundancy, monitoring, and automatic failover. If a server crashes, another takes over. You do not need to worry about power failures, OS updates, or hardware maintenance.
Self-hosted platforms require you to manage uptime. Our deployment uses a launchd service on macOS that automatically restarts the API server if it crashes. We have crash recovery that reconciles exchange state on restart. But if the machine loses power or network connectivity, trading stops until someone intervenes (or the dead man switch triggers after 24 hours and shuts everything down safely).
For most traders, this reliability trade-off favors self-hosting because the alternative (trusting a third party with your exchange credentials) is a larger risk than occasional downtime. A few hours of missed trading opportunities is recoverable. A credential breach is not.
Who Should Use What
Cloud platforms are appropriate for traders with small portfolios (under 5,000 dollars), limited technical ability, and a preference for simplicity over control. The convenience of a managed platform and the simplicity of pre-built strategies outweigh the security and cost concerns at small scale.
Self-hosted platforms are appropriate for traders with larger portfolios, technical comfort with running software (or willingness to learn), and a need for custom strategies, rigorous backtesting, or advanced risk management. The security, cost, and control advantages become significant as portfolio size increases.
We run self-hosted because we manage 45 bots across a 45,000 dollar paper portfolio with the intention of scaling to live trading. At that scale, the security and control requirements are non-negotiable. The operational overhead of managing our own infrastructure is trivial compared to the risk of trusting a third party with our exchange access.