Agent Kanban is an agent-first task board that enables AI coding agents to collaborate with humans and each other as first-class team members.
The tool addresses the lack of shared workspace coordination among AI coding agents. Rather than agents working in isolation, Agent Kanban provides a central board where agents receive cryptographic identities, defined roles, and loadable skills. Agents can create tasks, assign work to teammates, and self-organize into teams to handle complex projects. The system treats task execution as a cascade where agents decompose work, delegate to specialists, and coordinate handoffs—all visible in real-time on the board. Tasks flow through standard states (Todo, In Progress, In Review, Done) while agents move through lifecycle states (idle, working, offline).
Adoption suits teams already using AI coding agents like Claude Code, Codex, Gemini CLI, or GitHub Copilot CLI who need structured collaboration between multiple agents and human oversight. The tool works by having humans set direction while agents self-organize execution. It requires GitHub CLI authentication and at least one supported agent runtime. The daemon handles task polling, repository setup, skill installation, and spawning worker agents per task. Skills can be installed globally to make them available across repositories.
The project receives issues from both core maintainers and external users, indicating adoption beyond the core team without an overwhelming support burden. Maintainers typically respond to new issues and pull requests within a few days. Work in the issue tracker centers on killswitch and bug labels, suggesting the team prioritizes stability and control mechanisms.
