Daintree is an open-source project that functions as a delegation environment for orchestrating AI coding agents. It allows users to manage sessions for various AI models, including Claude, Gemini, and Codex, across different Git worktrees. The platform integrates terminals, facilitates context injection, and automates workflows to improve the efficiency of AI-assisted development.
The Daintree Assistant, a key component, operates on top of existing agent CLIs (e.g., Claude Code, Gemini CLI, Codex, GitHub Copilot CLI). It can spawn new agent terminals in any worktree, broadcast prompts to multiple agents simultaneously, monitor their progress, and execute Git operations. The assistant is designed to trigger any action available in Daintree's action palette, functioning as a sandboxed agent session that connects to a local Daintree MCP server for action system exposure and a Daintree docs server for answering 'how-to' questions.
Recent updates in version v0.13.0 include a reworked panel grid for improved scrolling with large fleets of agents, an enhanced action palette with pinning and scope chips, and improvements to the Daintree Assistant for surfacing launch failures and MCP cascade details. Performance enhancements focus on cold start optimization, deferring PTY and workspace host forks, and overlapping IPC during the first render. The project is written in TypeScript and is available for macOS, Windows, and Linux.