Descriptions:
Developer “bri” builds a fully autonomous cryptocurrency trading bot on the Solana blockchain using OpenClaw — an open-source AI assistant framework backed by Anthropic’s Claude — and controls it entirely through Telegram messages. The video continues from a prior installation guide in which OpenClaw was securely deployed on a Google Cloud Platform VM to isolate it from personal data on the host machine.
The build process is driven entirely by natural language: trading parameters are typed into Telegram and the Claude-powered agent handles implementation. The bot is configured for momentum trading on established tokens with at least $1M liquidity and 24-hour trading history, given a $500 starting budget, and instructed to use Jupiter as its DEX aggregator for optimal swap prices. The agent creates its own Solana wallet, runs as a persistent daemon process with 10-second market checks, Docker-based crash recovery, 5-minute heartbeat logs, and Telegram alerts for key events. A trading.md learning journal is added so the bot documents its trades and refines its strategy over time.
What makes this demonstration notable is how little direct code is written — the entire bot architecture, wallet creation, and daemon configuration are generated through conversational prompts to the Claude-backed agent. The video makes a practical case that tools like OpenClaw can abstract away blockchain development complexity, making autonomous trading agents accessible to non-developers while the GCP sandbox approach addresses the security risks of giving an AI agent broad system access.
📺 Source: bri · Published February 08, 2026
🏷️ Format: Hands On Build
