Kander CLI
A terminal kanban board that drives several AI agents in parallel with worktree isolation and independent review.
- 172 ★ on the repository
- Last commit
Broad access
Why this level
- Launches several agents that edit code and use Git without your involvement at each step.
- Reuses the credentials of the installed `gh` to import GitHub tasks.
Install
- Works in
Manual install
brew install dualface/tap/kanderFor macOS. On Linux download the binary from GitHub releases, on Windows take `kander-windows-amd64.zip`.
Access and checks
How we reviewBroad accessThe tool can change data, run commands or handle secrets.
- Launches several agents that edit code and use Git without your involvement at each step.
- Reuses the credentials of the installed `gh` to import GitHub tasks.
This is third-party code. Review the repository files before installing.
What it does
Kander is a terminal task board for several coding agents. Each task gets its own Git worktree and a tmux or herdr terminal container. The scheduler checks affected files and branches in advance so parallel agents do not write to the same place. After the executor, an independent reviewer checks the task in two stages, and delivery does not close until review passes. Tasks come from GitHub Issues through an already configured `gh`; Kander stores no tokens. It works with Codex, Claude, Grok, Cursor and Pi.
Who it is for. Developers running half a dozen agents at once who want task state on one board rather than in scattered terminals.
Good fit when
- Several agents edit one repository and write conflicts must be avoided.
- An agent's result must pass independent review before merge.
- GitHub Issues should be imported with one key press, `g`.
- A card's work is handed from one agent to another.
Not a fit when
- One agent and one task; the board adds extra steps.
- Git and none of the supported agents are installed.
- You need a board for people rather than for agents.
Example request
Create a task to refactor the payment module, launch it through the kanban flow and wait for independent review.Limitations
It needs Git and at least one agent from Codex, Claude, Grok, Cursor or Pi. macOS installs through Homebrew, Linux downloads a binary, Windows takes an archive from releases. The first launch opens a rules setup wizard, and `kander doctor` checks and repairs the environment. The load figures in the README (2,477 cards over 90 days) were measured by the author on his own projects. The last commit is from November 2026. Importing from GitHub needs an installed `gh` that is logged in. Agents perform the review, so it also spends tokens.
How to disable. Remove the binary (`brew uninstall kander` on macOS) and the connected rules from agent configuration. Task worktrees and branches stay in the repository and can be deleted with Git.
FAQ
How do I install Kander?
On macOS: `brew install dualface/tap/kander`, then `kander`. On Linux download `kander-linux-<arch>.tar.gz` from the latest release; on Windows take `kander-windows-amd64.zip`.
How do I hand a card to another agent?
Stop the current agent, launch a new one and ask it to continue card `TASK-ID`. The `y` key on the board copies the identifier.
Is a GitHub token required?
No. Kander uses the credentials of the installed `gh` and neither reads nor stores the token itself.
What does the review check?
Two independent review stages. PMQA catches semantic and state-machine bugs. A task does not close while findings remain.
What does `kander doctor` do?
It checks dependencies, agent availability, launchers and rules configuration, then repairs what can be repaired automatically.
Discussion
No questions or reviews yet.
Related
CLI toolsA workspace for parallel Claude Code and Codex tasks, with separate Git worktrees, terminals and change review
Four skills for agents working with the Agent Kanban task board: worker, maintainer, task creator and project planner, with human review of the result
A local control panel for coding agents: it runs Claude Code and Codex, keeps a task queue and merges verified work into main.
A Rust terminal task board: every task runs in its own git worktree and tmux window while Claude Code, Codex, Gemini CLI and other agents work in parallel
A planner for Claude Code, Codex and OpenCode: one goal becomes a task plan and independent tasks run in parallel
DevChain runs a team of AI coding agents on your own machine with a shared board, chat, code review and live terminals.
README in short
A Go project with the `kander` and `kander doctor` commands. Docs cover card transactions, coordinator recovery and GitHub issue import. READMEs exist in English, Chinese and Japanese.