A complete guide to installing and using OpenCode — the open-source, terminal-native AI coding assistant that supports 75+ LLM providers as a powerful alternative to Claude Code.
If you have been using Claude Code and love the terminal-based AI coding experience but wish you had more flexibility — different models, open-source transparency, and no vendor lock-in — OpenCode is worth your attention.
Built by the SST team, OpenCode is an open-source, terminal-native AI coding assistant that supports 75+ LLM providers including Anthropic, OpenAI, Google, and local models via Ollama. This guide walks you through installation, configuration, and daily usage.
OpenCode is a CLI/TUI (Terminal User Interface) tool that brings agentic AI coding to your terminal. Key differences from Claude Code:
| Feature | Claude Code | OpenCode | | ---------------- | ------------------ | ---------------------------- | | Source | Proprietary | MIT open-source | | Models | Claude only | 75+ providers + Ollama | | Pricing | $20-200/mo or API | Free tool, pay your provider | | Architecture | CLI tool | Client/server + HTTP API | | Local Models | No | Yes (Ollama native) | | Desktop App | Research preview | Beta available | | IDE Support | VS Code, JetBrains | VS Code |
The biggest advantage is provider independence. You can use Claude, GPT, Gemini, Llama, or any other model — and switch between them freely.
The fastest way to install OpenCode:
curl -fsSL https://opencode.ai/install | bash
This downloads the latest binary and adds it to your PATH.
brew install sst/tap/opencode
If you have Go installed:
go install github.com/sst/opencode@latest
git clone https://github.com/sst/opencode.git cd opencode go build -o opencode . mv opencode /usr/local/bin/
opencode --version
When you first run OpenCode in a project directory, it will guide you through setup:
cd your-project opencode
OpenCode supports multiple authentication methods:
# Set your API key export ANTHROPIC_API_KEY=sk-ant-your-key-here # Or use OAuth (if available) opencode auth login anthropic
export OPENAI_API_KEY=sk-your-key-here
export GOOGLE_API_KEY=your-key-here
Run models entirely on your machine with zero API costs:
# Install Ollama curl -fsSL https://ollama.com/install.sh | sh # Pull a coding model ollama pull qwen2.5-coder:32b # Configure OpenCode to use it opencode --provider ollama --model qwen2.5-coder:32b
OpenCode stores configuration in opencode.json at your project root:
{ "$schema": "https://opencode.ai/config.json", "provider": "anthropic", "model": "claude-4-5-sonnet-20260101", "mcpServers": { "filesystem": { "command": "npx", "args": ["-y", "@modelcontextprotocol/server-filesystem", "."] } } }
Once inside the TUI, you interact with OpenCode conversationally:
> Fix the TypeScript error in src/utils/auth.ts
> Write unit tests for the UserService class
> Refactor the database queries to use connection pooling
> Explain what this function does and suggest improvements
| Shortcut | Action |
| -------- | ------------------------ |
| Enter | Send message |
| Ctrl+C | Cancel current operation |
| Ctrl+L | Clear conversation |
| Ctrl+O | Open file picker |
| Tab | Autocomplete file paths |
| Esc | Exit OpenCode |
OpenCode can read, write, and edit files directly:
> Read the file src/app/page.tsx and explain the routing logic
> Create a new API endpoint at src/app/api/users/route.ts
that handles GET and POST requests
> Update the README.md with installation instructions
OpenCode can execute terminal commands on your behalf:
> Run the test suite and fix any failures
> Install the zod package and add schema validation
to the API routes
> Start the dev server and check for compilation errors
One of OpenCode's powerful features is MCP support. You can connect external tools and data sources:
{ "mcpServers": { "github": { "command": "npx", "args": ["-y", "@modelcontextprotocol/server-github"], "env": { "GITHUB_TOKEN": "ghp_your_token" } }, "postgres": { "command": "npx", "args": ["-y", "@modelcontextprotocol/server-postgres"], "env": { "DATABASE_URL": "postgresql://user:pass@localhost:5432/db" } } } }
OpenCode uses per-agent glob patterns for MCP, meaning you can control which tools are available to which conversations — giving you more granular control than Claude Code's approach.
Switch models based on the task:
# Complex architecture decisions — use the strongest model opencode --model claude-4-5-sonnet-20260101 # Quick fixes and simple tasks — use a faster/cheaper model opencode --model claude-4-5-haiku-20260101 # Privacy-sensitive code — use a local model opencode --provider ollama --model qwen2.5-coder:32b
Tailor OpenCode's behavior per project by adding instructions to your config:
{ "instructions": "This is a Next.js 16 project using TypeScript, Tailwind CSS, and Payload CMS. Always use server components by default. Follow the existing code style." }
OpenCode's client/server architecture means sessions can persist even when you close the terminal. The SST team is building toward Workspaces that survive across sessions — a feature Claude Code's simpler CLI cannot support.
Since you bring your own API keys, monitor usage:
# Check token usage for the current session # OpenCode shows token count in the TUI footer
Choose Claude Code if:
Choose OpenCode if:
OpenCode brings the power of agentic AI coding to your terminal without locking you into a single provider. Its open-source nature, 75+ model support, and client/server architecture make it a compelling choice for developers who value flexibility and control.
Install it, point it at your preferred LLM provider, and start coding. The learning curve is minimal if you have used Claude Code before — the core experience is familiar, just with more options.
