Player not loading? Watch on YouTube
This tutorial introduces OpenCode as a coding assistant that can switch between model providers within one terminal interface. The speaker says it supports subscription accounts, API keys and local models. The demonstrated connections use hosted providers; the tutorial does not show a local LLM setup or establish offline operation.
Setup starts with a curl installation command and a fresh terminal session if the command is not found. The speaker then opens OpenCode inside a project directory, including through the VS Code terminal. Provider examples cover OpenCode Zen, models labeled free, OpenCode Go and authentication with an existing ChatGPT subscription.
A practical example asks OpenCode to scaffold a NestJS server, then checks its Hello World response at localhost:3000. The walkthrough also demonstrates effort variants, separate sessions for refactoring and project rules, share links that can be revoked, and Markdown exports. The timeline offers options to undo message and file changes or fork a session from an earlier message. Shift+Tab switches between planning and build execution.
The later sections install a skill found through skills.sh, restart OpenCode to load it, and use /init to create or update AGENTS.md. Although the chapter names MCPs, the detailed example focuses on skills. The final workflow uses a personal knowledge base and an advisory skill built from YouTube transcripts to request business recommendations.
