The Automation of Task Management

The traditional Kanban board is evolving from a passive tracking tool into an active execution engine. Kanbots introduces a desktop application designed to run parallel agents on every card, transforming the board into a live environment where decisions surface and costs accrue as runs progress.

The architecture relies on dispatching agents—such as Claude Code or Codex—into individual git worktrees on specific branches. This allows for a high degree of concurrency; you can dispatch agents on as many cards as you want. The system is built on a local-first principle with 0 telemetry, ensuring that 0 bytes leave your machine.

The operational mechanics are centered on autonomy. Through an "autopilot" feature, personas can split work, run in parallel, and check their own output. This capability is reflected in the live board state, which tracks active tasks and progress. For example, a board might show 10 issues, with 2 active and 1 awaiting. The system even tracks granular cost analytics, such as $0.1318 or $0.3114, providing visibility into the financial impact of agentic workflows.

The software is available under an MIT license and is free forever, supported by pay-what-you-can donations. For those requiring a different scale of collaboration, a hosted version is available for teams at $19per seat / month or $190 billed yearly.

This shift toward agentic task management suggests a future where the bottleneck is no longer the execution of the task, but the quality of the prompt and the orchestration of the agents. As developers move from managing tickets to managing autonomous worktrees, the role of the engineer shifts toward high-level oversight and architectural decision-making.

Consider this: If your task management system can execute its own pull requests, what becomes the new primary metric for developer productivity?