What is AgentKan?

AgentKan is a desktop app that lets you manage multiple Claude Code agents through a familiar KanBan board interface. Instead of running agents one at a time in a terminal, you get a visual project management layer where your specialist agents pick up tasks and work on your codebase concurrently. Frontend devs, backend devs, QA engineers, or your own custom team members all contribute to your code base, managed visually with a kanban interface.

Think of it as a project manager for your AI dev team.

Features

Chat with your AI Project Manager

Describe what you want to build in plain language. The PM agent breaks your request down into tasks, assigns them to the right specialists, and manages the board for you.

Watch the PM take over

Once the PM understands your request, it creates tasks, assigns agents, and kicks off work — all automatically.

Drag-and-drop KanBan board

Track tasks as they move through Backlog, To Do, In Progress, Blocked, Review, and Done. Drag tasks between columns, or let agents move them automatically as they work.

Create tasks manually

Need to add a specific task? Create one directly with a title, description, priority, model selection, and assign it to the appropriate agent team member.

Review agent progress

Click into any task to see what the agent has done — including the full activity log, cost tracking, and file changes.

Meet your team

A roster of specialist agents — frontend, backend, DevOps, QA, and more — each with a unique skill set and perspective. Define your own custom team members with unique "personalities" and define what skills and tools they have available.

Run multiple projects at once

Manage several codebases simultaneously. Each project gets its own board, agents, and cost tracking.

Slack integration

Chat with your PM directly from Slack. Get updates, give instructions, and manage your board without leaving your workflow.

Git worktree

Each agent operates in their own git worktree, reducing the risk of code collisions.

curl -fsSL https://raw.githubusercontent.com/tencubed/agentkan-release/main/fix-gatekeeper.sh | bash

How it works

AgentKan runs entirely on your machine. The app bundles a local backend server that communicates with Claude via the Anthropic API. Your code never leaves your machine — only the prompts and agent conversations go through the API.

Each agent runs as an independent Claude Code session pointed at your project directory. The KanBan board coordinates task assignment, tracks progress in real time via WebSocket, and lets you review everything agents have done.

Security and Privacy

No datalogging, no remote tracking, no analytics. The only outbound call is to GitHub to check for new updates and releases.

Built with

• Claude Agent SDK — powers the AI agents
• Electron — native desktop app
• React + Material UI — frontend
• SQLite + Drizzle ORM — local database