Kanban CLI gives coding agents a local task board humans can review
Kanban CLI is a local-first, Rust-based project management tool for coordinating AI coding agents with schema-validated tasks, Git worktrees, and human review gates.
Kanban CLI is a terminal-based, local-first task manager built for agent-driven software workflows. Its repository describes a Rust CLI that stores structured project data locally, integrates with Git, and uses JSON-backed task, role, sprint, and agent concepts. The matching Show HN post frames it as a coordination layer for reducing agent randomness by forcing work through schema-validated tasks, isolated worktrees, and human review.
Key takeaways
- Kanban CLI targets developer workflows where AI agents need explicit task boundaries, acceptance criteria, and reviewable state.
- The repository describes local file-system storage, Git integration, role assignments, sprint/task commands, and documentation for JSON output.
- The Show HN post says the intended loop is: read a skill, receive a strict JSON payload, implement in an isolated Git worktree, then wait for a reviewer to move the task to done.
- The project is active, with recent commits around Rust performance, schema self-healing, and task-field detection.
- It is licensed under AGPL-3.0, so teams should review license fit before embedding it into internal or commercial tooling.
Practical LinkLoot angle
Kanban CLI is interesting because it treats agent work as project-management state, not just chat history. That is useful when several coding agents or long-running sessions need the same source of truth: what files should change, which role owns the task, what acceptance criteria matter, and whether a human has approved the merge.
| Tool or workflow | Best use | Limitation to check | Source |
|---|---|---|---|
| Kanban CLI | Local task boards for agentic coding sessions | Early project; verify commands and schemas before relying on it | Codeberg repository |
| Plain GitHub issues | Team-visible planning and review | Less convenient for offline/local agent loops | GitHub workflow comparison |
| tmux plus notes | Fast local coordination | Easy for agents to drift without structured state | Practical workflow comparison |
A practical test is to run it on a non-critical repository and ask an agent to complete one small issue with explicit files_to_modify, acceptance criteria, and a human-only final merge. If the task state stays readable after the agent fails halfway through, the tool is solving a real coordination problem rather than just adding another dashboard.
What to verify before you act
Check the AGPL-3.0 license before using Kanban CLI inside proprietary delivery workflows. Review the JSON schemas and Git worktree behavior against your own repository hygiene rules, especially if agents can write to terminals or branches automatically. Because the HN launch is recent and the project is still moving quickly, pin a version or commit before building repeatable automation around it.
It is a local-first terminal project-management tool designed to coordinate AI coding agents with structured tasks, roles, sprints, and Git-integrated workflows.
For more agent workflow ideas, see LinkLoot's guide to AI agent tools.