A local-first, extensible Kanban board designed for orchestrating AI agents. It allows users to manage tasks as AI agent workflows, supporting parallel execution, automated splitting, and self-checking of tasks. Ideal for developers and researchers seeking to integrate and manage multiple AI coding and task-completion agents within a unified interface, streamlining complex AI-driven development processes.

How It Works

Kanbots operates by dispatching AI agent CLIs (like Claude Code, Codex, Gemini) to execute tasks within isolated, per-run worktrees. Each worktree is branched from the repository's default branch, ensuring task isolation. Agent output streams directly to the UI in real-time, with user decisions prompted interactively. The system supports local SQLite-based issue tracking or direct integration with GitHub Issues, mirroring status changes. An "Autopilot" mode enables sophisticated, multi-persona workflows, including parallel task splitting and automated QA cycles.

Quick Start & Requirements

• Primary install: npx kanbots
• Prerequisites: Node.js (20+), pnpm (10+), git, and at least one AI agent CLI (e.g., claude, codex, gemini) configured on the system PATH. GitHub CLI (gh) required for GitHub integration.
• Resource footprint: Agent binaries are downloaded on first run (~80MB).
• Links: Official Install Script (macOS), Releases Page, Getting Started Docs.

Highlighted Details

• Supports 11+ AI agent CLIs, including Claude Code, Codex, Gemini, and Copilot, plus any ACP-compatible CLI.
• "Autopilot" mode orchestrates complex workflows: multi-persona task splitting, parallel execution slots, and automated Quality Assurance (QA) cycles.
• Dual-mode issue management: local-first SQLite database or direct integration with GitHub Issues, mirroring status changes.
• "Branch preview" allows launching development servers from agent worktrees; "Promote" action lands agent work as commits or draft PRs.
• Model Context Protocol (MCP) server enables integration with MCP-aware applications like Cursor and Claude Desktop.

Maintenance & Community

No specific details regarding maintainers, community channels (e.g., Discord, Slack), or project roadmap were found in the provided README.

Licensing & Compatibility

• License type: MIT.
• Compatibility: The MIT license permits commercial use and integration with closed-source projects.

Limitations & Caveats

Packaged macOS builds are unsigned, requiring manual intervention or a specific install script to bypass Gatekeeper. Windows builds are unsigned, triggering Microsoft Defender SmartScreen warnings. Users must pre-install and authenticate individual AI agent CLIs on their system PATH.