The Complete OpenCode Tutorial: Keep Claude Code's Playbook (Skills, MCP, agents.md) While Freely Choosing Your Model

Most coding agent tools lock you into one provider, one model, one agent -- switching agents usually means switching tools entirely. OpenCode, the subject of this tutorial, solves that: with a single AI agent, you can freely use your existing subscriptions, free models, API-key-backed models, or even local models. And critically, you keep the exact features that Claude Code and Codex popularized -- plan/build mode, skills, MCPs, agents.md files, sub-agents, and workflows -- all inside the same tool.

The walkthrough highlights practical UI touches -- changed code highlighted in green, easy navigation between sessions -- alongside one particularly useful feature: the `share` command, which converts an entire conversation history into a shareable HTML page. Opening the generated URL in a browser shows the whole session, from the initial prompt through everything the agent generated. Running `unshare` instantly kills that URL, making it inaccessible again.

In the back half of the video, the creator reveals his own personal workflow for using OpenCode as a 'second brain.' For decisions -- like how to grow a community -- he triggers pre-built skills to pull in feedback from several distinct perspectives at once, essentially convening a virtual board of advisors. The tutorial covers the full path from installation through skills, MCPs, and agents.md basics, all the way to this kind of applied, real-world use case.