Player not loading? Watch on YouTube
This tutorial walks through an OpenHands deployment on Elestio and a coding task using GPT-5. The presenter describes OpenHands as an open source, MIT-licensed platform that runs coding agents in isolated sandboxes on a machine or in the cloud. Although the introduction mentions local model support, the demonstration uses an OpenAI API key rather than a local LLM.
Setup covers choosing a cloud provider, region and service plan, opening the deployed instance, and configuring the model provider. Existing projects can connect through GitHub, GitLab or Bitbucket. For the demonstration, the presenter starts a new conversation and asks the agent to build a procedural Crossy Road-style game with Three.js.
The agent creates a task list and project files. The presenter inspects the workspace through VS Code after the built-in code view fails to open as expected, then checks terminal activity and the app preview. The game runs, but the presenter finds the traffic too dense and says it needs further iteration.
The remaining overview covers mobile access to the self-hosted instance, cost and context displays, custom microagent skills, manual model names, memory settings and optional confirmation mode. It also introduces MCP connections, conversation budget limits and documentation examples for automated code review and dependency upgrades.
