Summary

This project provides a UI for managing AI agent teams, positioning the user as a CTO overseeing autonomous agents that handle tasks, communicate, and review code. It aims to simplify complex AI workflows by offering a local, zero-configuration solution where agents operate autonomously on a Kanban board, reducing the need for direct user intervention. The primary benefit is enabling sophisticated AI-driven task management and development cycles without API keys or complex setup, suitable for users who want to delegate work to AI teams.

How It Works

The core approach involves creating agent teams with distinct roles that work autonomously and in parallel. Agents communicate via a real-time messaging system, manage their own tasks, and implement a built-in review workflow where they critique each other's code. The UI visualizes this process on a Kanban board, allowing users to monitor task status, review code changes at a hunk level, inspect tool usage, and trace task-specific logs. Users can intervene by sending direct messages or selecting quick actions, offering flexible autonomy from fully autonomous operation to per-action approval.

Quick Start & Requirements

• Primary install/run command: Download the app for your platform. On first run, a setup wizard handles installation and authentication of Claude Code.
• Non-default prerequisites: Requires a Claude Code plan from Anthropic to run agents. May trigger OS security warnings (e.g., SmartScreen) requiring manual override. Runs entirely locally; no cloud backend.
• Links: No external quick-start or demo links provided in the README.

Highlighted Details

• Zero-setup onboarding with built-in Claude Code installation and authentication.
• Task creation supports attachments (images, files), with agents automatically incorporating them.
• Deep session analysis provides a 6-category token tracking breakdown for cost monitoring.
• Built-in code editor with Git support allows file editing without leaving the application.
• Flexible autonomy settings enable granular control over agent actions, from full autonomy to per-action approval.
• Post-compact context recovery ensures consistent team-management context is preserved.

Maintenance & Community

The project includes a roadmap detailing planned features such as planning mode, a visual workflow editor, multi-model support, and remote agent execution. Contributing guidelines and a Code of Conduct are available. No specific community channels (like Discord/Slack) or notable contributors/sponsorships are mentioned.

Licensing & Compatibility

• License type: AGPL-3.0.
• Compatibility notes: The AGPL-3.0 license is a strong copyleft license. It requires any derivative works or software that links to it to also be licensed under AGPL-3.0, which may impose restrictions on commercial use or integration with closed-source systems.

Limitations & Caveats

The "Multi-agent backend" is noted as "In development," suggesting potential limitations or ongoing changes. The AGPL-3.0 license carries significant implications for commercial adoption and integration with proprietary software due to its strict copyleft requirements.