Opinionated AI coding agent for agentic workflows.
Based on OpenCode by Anomaly Co (MIT License)
A lightweight fork of OpenCode optimized for single developers running agentic workflows. Think PM agents orchestrating specialist agents, with deep memory/context integration via remory.
Philosophy:
- Opinionated defaults over endless configuration
- Optimized for agentic orchestration patterns
- Lightweight - stripped of enterprise bloat
- Single developer friendly
# Clone
git clone https://github.com/randomm/opencode.git
cd opencode
# Install dependencies
bun install
# Build
cd packages/opencode && bun run build
# Install binary (adjust path for your platform)
cp dist/opencode-darwin-arm64/bin/opencode ~/bin/
# Linux x64
# cp dist/opencode-linux-x64/bin/opencode ~/bin/Make sure ~/bin is in your PATH.
# Navigate to your project
cd my-project
# Start opencode
opencodeOn first run, you'll be prompted to configure your AI provider. Set your API key:
export ANTHROPIC_API_KEY="sk-..."
# or
export OPENAI_API_KEY="sk-..."Create opencode.json in your project root or ~/.config/opencode/config.json for global settings.
{
"provider": {
"anthropic": {
"model": "claude-sonnet-4-20250514"
}
}
}Key options:
provider- AI provider configuration (anthropic, openai, etc.)mcpServers- MCP server integrationsinstructions- Custom system instructions
Two built-in agents, switchable with Tab:
- build - Full access agent for development work (default)
- plan - Read-only agent for analysis and exploration
Use @general in messages to invoke the subagent for complex searches.
This fork focuses on:
- Agentic workflows - Optimized for PM/specialist agent patterns
- remory integration - Deep memory and context management
- Minimal footprint - No desktop app, no VS Code extension, no npm publishing
- Opinionated defaults - Less configuration, more convention
Removed from upstream:
- Desktop application
- VS Code extension
- npm/brew/scoop publishing
- Enterprise features
MIT License. See LICENSE.
This project is a fork of OpenCode by Anomaly Co, licensed under MIT. We're grateful for their work on the original project.