Skip to content
Elmer Augusto Jacobo Otiniano
ESEN — View in Spanish
Back to blog
AI & AgentsClaude CodeOpenCodeMCPAgentsTerminal3 min read

Claude Code vs OpenCode: which one to choose for your workflow

I compare Claude Code and OpenCode on installation, models, cost and configuration, and explain when I use each one depending on the project.

I use Claude Code and OpenCode depending on the project. I choose based on what I need in each case: getting started fast or having control.

Here is the comparison, based on what each tool's official docs say.

What each one is

Claude Code is Anthropic's agentic coding tool. It reads your codebase, edits files, runs commands and connects to your development tools. It works in the terminal, VS Code, JetBrains, a desktop app and the web.

OpenCode is an open source coding agent. It also has a terminal interface (TUI), a desktop app (in beta) and an IDE extension. The sst/opencode repository uses the MIT license.

Installation

Both install with one command:

# Claude Code
curl -fsSL https://claude.ai/install.sh | bash

# OpenCode
curl -fsSL https://opencode.ai/install | bash

If you prefer Homebrew:

brew install --cask claude-code
brew install anomalyco/tap/opencode

After that it's claude or opencode inside your project. With Claude Code you log in the first time. With OpenCode you connect a provider with /connect.

Models and cost

Claude Code uses Anthropic models. You need a Claude subscription or a Console account. Pro costs 20 USD a month (17 USD if you pay annually) and Max starts at 100 USD a month. The terminal CLI, VS Code and JetBrains also accept third-party providers.

OpenCode doesn't tie you to one provider. Its documentation mentions 75+ providers: you add your API key and that's it. It also supports local models through Ollama, or any OpenAI-compatible endpoint. If you want to try it without deciding yet, there's OpenCode Zen, a set of models the team has already tested.

Configuration and extension

Both read a project instructions file, and the two are not that different.

In Claude Code it's CLAUDE.md. If your repo already has an AGENTS.md, it can read that too. On top of that you get skills to package repeatable workflows, hooks that run commands before or after an action, subagents for parallel work, and an Agent SDK if you want to build your own agents.

In OpenCode, /init creates an AGENTS.md at the project root. It ships with two built-in agents you switch with Tab: build, with full access, and plan, which is read-only, doesn't edit files and asks permission before running bash. It also has /undo, /redo and /share, and you can call a general subagent with @general.

Both support MCP. To see how it's set up in OpenCode, I covered it in how to connect Figma to OpenCode using MCP.

When I choose each one

My rule is simple:

  • Claude Code when I want speed. I install it, log in and start working. I don't have to pick a provider or a model before writing the first instruction.
  • OpenCode when I want control. I choose the model, I can switch providers without switching tools, and it lets me customize the configuration in depth.

They don't compete. Since both can read AGENTS.md, your project instructions work in either.

What to do today

Install the one you don't use on a small project and give it the same task you already solved with the other. That's where you notice whether speed or control matters more to you.