Your personal AI cofounder for life.

A living AI agent system built on OpenCode that makes you 10x more effective. Stop jumping between tools and contexts — build a system that grows with you.

• Persistent Memory — Short-term (weekly focus) and long-term (preferences, patterns, decisions) that grows with you
• Workspaces — Your work areas (dev projects, business, personal) with dedicated context
• Smart Commands — /capture for quick ideas, /remember for important notes, /summarize for session documentation
• Specialized Agents — 10+ specialized agents (development, frontend design, research, code review, onboarding) — all ready to use
• Skills — Built-in expertise on OpenCode and other tools
• Fully Modular — Customize agents, commands, skills — make it yours

The problem: You use Claude Code, ChatGPT, OpenCode and Cursor - your context is scattered across different tools. Every time you start a new conversation, you're starting from zero.

The solution: OpenAgent gives you a persistent, personalized AI system that knows who you are, what you're working on, how you work and grows with you. It's your base system that you build up from.

Unlike other tools that overwhelm you with thousands of app connections and configuration options, OpenAgent focuses on what matters: getting things done. It's simple, modular and customizable to how you actually work.

• OpenCode installed
• Git

One-line install (recommended):

cd ~ && curl -fsSL https://open-agent.sh/install | bash

This will:

• Download OpenAgent to ~/openagent
• Run interactive setup
• Create symlinks to OpenCode config

Windows (PowerShell):

Windows users: use the Windows install skill at system/skills/windows-install-skill/SKILL.md. Clone this repo (or copy just that skill into your coding agent tool) and ask the agent to follow the steps. The Unix one-liner curl -fsSL ... | bash only works in WSL or Git Bash.

Manual install (for contributors):

# Clone the repo
cd ~ && git clone https://github.com/openagent-sh/openagent
cd openagent

# Run setup
./system/scripts/setup.sh

# 1. (Optional) Add shell alias for easy launching
# For zsh (most macOS users):
echo 'alias openagent="cd $OPENAGENT_ROOT && opencode --agent openagent --prompt Hello"' >> ~/.zshrc && source ~/.zshrc

# For bash:
echo 'alias openagent="cd $OPENAGENT_ROOT && opencode --agent openagent --prompt Hello"' >> ~/.bashrc && source ~/.bashrc

# If you get API errors, add --model flag:
# alias openagent="cd $OPENAGENT_ROOT && opencode --agent openagent --model anthropic/claude-sonnet-4-5 --prompt Hello"

# 2. Start OpenCode
opencode

# 3. Personalize your system
/onboarding

# 4. Start using OpenAgent
/openagent

# Capture a quick idea
/capture add dark mode to the dashboard

# Save something to remember
/remember stop running the dev server!!!!

# End your session with a summary for future references
/summarize 

After setup, run /onboarding to personalize your system (takes 5-10 minutes). The more information you share, the better your system will be set up

The onboarding flow will ask about:

• Your name (optional)
• Current projects and work areas
• Workspaces to create (dev, business, personal, etc.)
• Preferences and communication style
• This week's focus

After onboarding, you'll have:

• ✅ Personalized AGENT.md with your context
• ✅ Workspaces set up for your areas
• ✅ Long-term memory initialized
• ✅ This week's focus ready to go

Note: Technical setup (symlinks, agents, commands) is handled by setup.sh before onboarding.

See full onboarding flow

┌─────────────────────────────────────────────────────────────────┐
│                         OPENAGENT                               │
│                  Personal AI Cofounder System                   │
└─────────────────────────────────────────────────────────────────┘
                                │
                                ▼
┌─────────────────────────────────────────────────────────────────┐
│                        OPENCODE CLI                             │
│  Terminal TUI • Agents • Commands • Skills • Plugins            │
└─────────────────────────────────────────────────────────────────┘
                                │
                                ▼
┌─────────────────────────────────────────────────────────────────┐
│                      CONTEXT LAYER                              │
├────────────────┬────────────────┬────────────────┬──────────────┤
│   SOUL.md      │   AGENT.md     │    Memory      │    Tasks     │
│  (Identity)    │    (User)      │                │              │
│                │                │  ┌──────────┐  │  ┌────────┐  │
│  Core values   │  Your context  │  │MEMORY.md │  │  │ inbox/ │  │
│  How agent     │  Professional  │  │(long)    │  │  │backlog/│  │
│  operates      │  Preferences   │  ├──────────┤  │  └────────┘  │
│                │  Goals         │  │ short/   │  │              │
│                │                │  │(weekly)  │  │              │
│                │                │  ├──────────┤  │              │
│                │                │  │sessions/ │  │              │
│                │                │  │(logs)    │  │              │
│                │                │  └──────────┘  │              │
└────────────────┴────────────────┴────────────────┴──────────────┘
                                │
          ┌─────────────────────┼─────────────────────┐
          ▼                     ▼                     ▼
┌──────────────────┐  ┌──────────────────┐  ┌──────────────────┐
│     Agents       │  │     Skills       │  │   Workspaces     │
│                  │  │                  │  │                  │
│  openagent       │  │  opencode-skill  │  │  dev/            │
│  onboarding      │  │  frontend-design │  │  business/       │
│  capture-agent   │  │  skill-creator   │  │  personal/       │
│  developer-1     │  │  agent-browser   │  │  misc/           │
│  developer-2     │  │                  │  │                  │
│  frontend-design │  │                  │  │                  │
│  code-reviewer   │  │                  │  │                  │
│  opencode-agent  │  │                  │  │                  │
│  researcher      │  │                  │  │                  │
│  project-owner   │  │                  │  │                  │
└──────────────────┘  └──────────────────┘  └──────────────────┘
                                │
                                ▼
 ┌─────────────────────────────────────────────────────────────────┐
│                        COMMANDS                                 │
├──────────────────┬──────────────────┬──────────────────┬─────────┤
│ /openagent       │ /capture         │ /remember        │ /summarize │
│ Load full        │ Quick task       │ Save to          │ Session    │
│ context          │ to inbox         │ long-term        │ summary    │
├──────────────────┼──────────────────┼──────────────────┼─────────┤
│ /opencode-agent  │ /researcher      │ /developer-1     │ /project   │
│ System eng.      │ Tech research    │ Feature impl.    │ -owner     │
├──────────────────┴──────────────────┴──────────────────┴─────────┤
│ /frontend-designer                                            │
│ UI/UX polish                                                    │
└─────────────────────────────────────────────────────────────────┘

How it works:

1. /openagent loads your full context (SOUL + AGENT + Memory + Tasks)
2. Agents handle specialized work (dev, frontend design, onboarding, research)
3. Skills provide expertise (OpenCode customization, system documentation)
4. Commands capture ideas, save memories, summarize sessions, or spawn agents
5. Workspaces organize your projects by area

• OpenCode Docs — Official OpenCode documentation
• Sessions Template — Session summary format
• Memory System — How memory works
• Workspaces Guide — Organizing your work areas

Symlink Setup: OpenAgent uses symlinks to sync with OpenCode (~/.config/opencode/ → ~/openagent/system/). Changes are automatic both ways. Run ./system/scripts/setup.sh to set up.

All sorts of contributions and requests are welcome <3

It's a way simpler system that is more focused on being your personal cofounder for life, focusing on your actual tasks instead of building email and calendar automations.

It's meant to be customizable and modular from the ground up, because everyone works differently.

• Support for other CLIs (claude code, droid, ...)
• A webapp with an agent orchestration view, dashboard, a clean markdown editor that is built AI native
• Extensions like Vercel Agent Browser, AI SDK, Exa AI websearch, ...

OpenAgent checks for updates automatically when you run /openagent. When an update is available, you'll see a notification.

To update:

cd ~/openagent && ./system/scripts/update.sh

Your personal data (AGENT.md, memory, tasks, workspaces) is always preserved during updates.

MIT

Built with OpenCode — opencode.ai