canopy

# Daintree

**A habitat for your AI coding agents.**

Daintree is a desktop environment where multiple AI agents work side by side — isolated, observable, and under your control. Instead of juggling terminal windows and manually wiring context between tools, Daintree gives your agents a stable place to run while you focus on reviewing their work.

It works with any CLI agent — [Claude Code](https://docs.anthropic.com/en/docs/claude-code), [Gemini CLI](https://github.com/google-gemini/gemini-cli), [Codex](https://github.com/openai/codex), [OpenCode](https://github.com/anomalyco/opencode), [Gemini Agent](https://docs.gemini.com/agent), [Kiro CLI](https://kiro.dev/docs/cli/chat/) — and stays out of the way.

---

## Why Daintree

Running AI agents in parallel is becoming the standard workflow. But the tooling around it hasn't caught up. You end up with a dozen terminal tabs, no visibility into what each agent is doing, and no clean way to review or merge the results.

Daintree solves this by providing:

- **Automatic isolation** — Each task gets its own git worktree. Agents never collide.
- **Visibility at a glance** — See what every agent is doing, which ones need input, and what's changed across all branches.
- **Review-first workflows** — The bottleneck isn't generation speed, it's reviewing what agents produce. Daintree is built around making that fast.
- **Zero lock-in** — Your machine, your keys, your choice of agents. Daintree is agent-agnostic by design.

---

## Features

### Worktree Dashboard

View all git worktrees in a single dashboard with real-time status. Daintree auto-detects associated Pull Requests and Issues from branch names, shows commit-based summaries of changes, and manages dev server lifecycles per worktree.

### Agent Orchestration

Run multiple agents in a panel grid. Daintree tracks agent state automatically — `idle`, `working`, `waiting`, `completed`, `failed` — via output analysis. A notification strip alerts you the moment any agent need