KanBots: Kanban Boards Meet Parallel AI Agents for Software Development

KanBots has emerged with an intriguing proposition: a kanban board system designed specifically to orchestrate and manage parallel AI coding agents. The tool bridges the gap between traditional project management methodologies and the emerging capabilities of AI-assisted software development, creating a structured environment where AI agents can operate autonomously yet remain visible and controllable.

At its core, KanBots addresses a fundamental challenge in the current AI-assisted development landscape: the chaotic nature of AI agent interactions. Instead of treating AI assistants as simple chat interfaces, KanBots positions them as first-class citizens in a development workflow, complete with their own cards, worktrees, and decision points.

The system operates on a simple premise: drop a folder, get a board. Each card on the kanban board represents a task that can be dispatched to an AI agent—either Claude Code or Codex. What sets KanBots apart is its ability to run multiple agents in parallel, each operating in its own isolated git worktree on a separate branch. This approach prevents interference between concurrent AI-generated code while maintaining visibility into each agent's progress.

"The board updates live as runs progress, decisions surface, costs accrue," explains the project's documentation, highlighting the real-time nature of the system. Unlike traditional AI coding tools that operate in isolation, KanBots creates a persistent record of AI-generated work, making it possible to review, approve, and integrate AI contributions systematically.

One of the most compelling features is Autopilot mode, which allows developers to define personas—such as Product Manager, Senior Engineer, UX Designer, and others—and set a parallelism count. The system then round-robins through these personas, automatically splitting parent issues into subtasks and evolving the backlog as agents discover work. This creates a self-evolving development workflow where personas spawn additional personas to address emerging requirements.

The decision-making process deserves particular attention. Rather than allowing AI agents to make changes silently, KanBots implements a decision prompt system where agents pause and present numbered options. Developers can click their choice, use slash commands like /spec or /review, or even edit and resubmit. This human-in-the-loop approach ensures that AI-generated code aligns with the project's goals while maintaining developer control.

KanBots takes a local-first approach, storing everything in a .kanbots/ directory next to the repository. The system uses SQLite for its database, requires no cloud account, and collects no telemetry. This design choice positions KanBots as a tool for developers who value privacy and want to keep their code on their machines. The cost tracking features—per-run, per-card, and per-project rollups—add another layer of control, allowing developers to set budget caps and prevent unexpected charges.

The integration with existing AI coding tools is seamless. KanBots supports both Claude Code and Codex through an AgentCliAdapter, allowing developers to use their existing authentication and API keys. The system can drive real GitHub issues using a personal PAT (Personal Access Token) and can promote worktrees to commits or open draft PRs with a single click. A pre-push hook ensures that agents never publish changes without explicit approval.

For teams, KanBots offers a cloud version at $19 per seat per month ($190 billed yearly) that adds collaboration features like real-time presence on the board, assignment notifications, cross-device sync, and audit logs. The team version addresses the fundamental limitation of the desktop app: it requires another person or device to unlock collaborative features. "OSS is what one person does on their machine; Cloud is what a team does together," states the project philosophy.

The inclusion of an MCP (Model Context Protocol) server extends KanBots' utility beyond its own interface. The kanbots-mcp-server exposes the board over the Model Context Protocol, allowing tools like Cursor, Claude Desktop, or any MCP-aware application to drive it. This integration positions KanBots as a first-class tool within the broader AI development ecosystem.

What makes KanBots particularly interesting is its thoughtful approach to the human-AI collaboration problem. Rather than treating AI as a replacement for developers, it positions AI as a specialized team member with defined roles and responsibilities. The persona system allows developers to leverage different AI capabilities—architecture, user experience, testing, etc.—in a structured way.

The tool also acknowledges the reality of AI development costs. By implementing cost tracking and budget caps, KanBots helps developers manage the potentially expensive nature of running multiple AI agents simultaneously. This practical consideration reflects a mature understanding of both the capabilities and limitations of current AI systems.

KanBots enters a market that's rapidly evolving, with AI coding assistants becoming increasingly sophisticated. However, many existing tools focus primarily on the interaction between a single developer and an AI assistant. KanBots distinguishes itself by addressing the coordination of multiple AI agents and the integration of their output into a coherent development workflow.

The project's dual licensing approach—free for individual use with a pay-what-you-can model for the desktop app, and a traditional SaaS model for teams—demonstrates an understanding of different market segments. This strategy allows individual developers to adopt the tool without financial barriers while providing a clear path for commercial teams.

As AI continues to transform software development, tools like KanBots may become essential for managing the complexity of AI-assisted workflows. By providing structure, visibility, and control, KanBots addresses some of the fundamental challenges of working with AI coding agents at scale. The project's GitHub repository and active development suggest that this is just the beginning of what could become an important category of development tools.

For developers looking to integrate AI more deeply into their workflow while maintaining control over the process, KanBots offers a compelling approach. It doesn't promise to replace developers but rather to augment their capabilities with a structured system for managing AI contributions. As the line between human and AI-generated code continues to blur, tools that provide clarity and control will become increasingly valuable.