Open source Kanban desktop app that runs parallel agents on every card

An open source desktop Kanban application now supports deploying multiple AI agents in parallel on each card, enabling autonomous task management without relying on cloud services. This development allows individual users and teams to orchestrate AI-driven workflows locally, with features for cost tracking and integration with existing AI tools, making it a significant step for AI-enhanced project management.

The project, hosted on GitHub under an MIT license, enables users to drop a folder, generate a Kanban board, and assign AI agents—such as Claude or Codex—to each card. Each agent runs in its own git worktree, allowing parallel processing and live updates on the board. The app supports local-first operation, storing data in a SQLite database next to the repository, with no cloud account, telemetry, or HTTP server required.

It includes features like autopilot mode, where personas (e.g., product manager, developer, reviewer) can be set to automatically split work, evolve the backlog, and manage costs. Users can also manually dispatch agents, review decisions through a live thread, and manage costs with per-run and per-card analytics. The app also integrates with GitHub issues and draft PRs, allowing seamless collaboration with existing development workflows.

Why It Matters

This development matters because it provides a local, open source alternative for AI-assisted project management, reducing reliance on cloud services and enabling privacy-conscious workflows. It offers a scalable way for solo users or small teams to leverage AI agents for complex task orchestration, potentially transforming how project workflows are managed with AI support.

Background

Traditional Kanban tools focus on task visualization and basic collaboration. Recent advancements in AI have enabled automation and intelligent task handling, but most solutions rely on cloud-based services. This project bridges that gap by bringing AI agent orchestration into a local desktop environment, supporting tools like Claude and Codex, and emphasizing privacy and cost control. The initiative builds on prior developments in AI automation and open source project management tools, aiming to provide a flexible, privacy-respecting platform for developers and solo users.

“This app turns AI agents from a curiosity into a system of record, enabling autonomous work and better collaboration within a local environment.”

— Project developer

“By supporting existing CLI tools and local storage, the project ensures privacy and flexibility for individual users and small teams.”

— Open source contributor

What Remains Unclear

It is not yet clear how well the app scales for larger teams or complex projects, and how robust the AI agent coordination is in practice. Ongoing development and user feedback will determine its practical effectiveness.

What’s Next

Future updates are expected to include enhanced automation features, broader provider support, and more integrations with existing development workflows. Developers are encouraged to follow the GitHub repository for new releases and community contributions.

Key Questions

Can I use this app without an internet connection?

Yes, the app is designed to run locally with no need for internet access, storing all data on your machine.

What AI tools are supported?

The app supports Claude Code and Codex, with the ability to switch between them using your existing login or API keys.

Is this suitable for team collaboration?

Yes, it supports team workflows via GitHub integration and cloud features, but it is primarily optimized for local solo or small-team use.

How does cost management work?

The app provides live cost analytics with per-run, per-card, and per-project rollups, allowing users to set budgets and stop runs when limits are reached.

Source: Hacker News