Kanbalone

日本語

Kanbalone is a small local kanban app optimized for individuals who develop together with AI.

Turn remote issues into a local execution workspace for implementation with AI.

You can start it quickly with Docker and use it right away. In addition to a browser-based web UI for humans, it also provides a JSON API that is easy for scripts and AI agents to use.

Import issues from systems such as GitHub Issues, GitLab, and Redmine into Kanbalone, build out a local implementation body, and keep progress in comments. The remote tracker stays the source of truth while Kanbalone becomes the execution workspace for an individual developer and their AI agents.

Why Kanbalone

• Optimized for one person and one machine working with AI, without user or permission management.
• All data is stored locally.
• Import remote issues from GitHub, GitLab, or Redmine and work on them as local execution tickets.
• Multiple boards for different work categories.
• Tags, comments, and ticket dependencies: blockers and parent/child tickets.
• Keep remote title and body as reference while building a richer local implementation body.
• Push only selected comments back to the remote issue to keep work logs simple.
• Lightweight JSON API designed for automation and AI agents.
• Create the first board and start managing tasks immediately without extra setup.

Quick Start

Run the published Docker image:

docker run --rm \
  -p 3000:3000 \
  -v kanbalone-data:/app/data \
  ghcr.io/wamukat/kanbalone:v0.9.35

Open:

http://127.0.0.1:3000

User guide:

https://wamukat.github.io/kanbalone/

Use ghcr.io/wamukat/kanbalone:latest when you want the newest released image instead of a pinned version.

Docker Compose

Run the published image with Compose:

docker compose -f docker-compose.image.yml up

Change the host port:

KANBAN_PORT=3457 docker compose -f docker-compose.image.yml up

The app stores its SQLite database at:

/app/data/kanbalone.sqlite

The provided Compose file uses a Docker named volume, kanbalone-data, for persistence.

On Windows, use Docker Desktop or Rancher Desktop with WSL2 and keep the named volume setup.

Kanbalone does not include authentication. Do not expose it directly to untrusted networks.

Local Development

pnpm install
pnpm dev

Default URL:

http://127.0.0.1:3000

Remote provider sandbox for GitLab and Redmine:

docker compose -f docker-compose.remote-providers.yml up -d
pnpm sandbox:remote-providers

Kanbalone shows the remote issue import action only when at least one remote provider credential is configured. The import panel lists configured providers only.

See Remote provider sandbox.

Codex Skill

Kanbalone includes a Codex skill for API-only kanban operations:

mkdir -p "${CODEX_HOME:-$HOME/.codex}/skills"
cp -R skills/kanbalone-api "${CODEX_HOME:-$HOME/.codex}/skills/"

When you use only the Docker image, fetch the skill from the matching GitHub release tag to the host running Codex:

tmpdir=$(mktemp -d)
curl -L https://github.com/wamukat/kanbalone/archive/refs/tags/v0.9.35.tar.gz \
  | tar -xz -C "$tmpdir" kanbalone-0.9.35/skills/kanbalone-api
mkdir -p "${CODEX_HOME:-$HOME/.codex}/skills"
cp -R "$tmpdir"/kanbalone-0.9.35/skills/kanbalone-api "${CODEX_HOME:-$HOME/.codex}/skills/"
rm -rf "$tmpdir"

The skill runs from the host and talks to the Kanbalone HTTP API, for example http://127.0.0.1:3000.

Documentation

Published user guide:

• https://wamukat.github.io/kanbalone/

For users and API clients:

• User guide
• Data model and concepts
• AI API guide
• API examples
• OpenAPI

For developers and maintainers:

• Development
• Remote provider sandbox
• Docker image distribution
• Release process
• Performance tooling
• Dialog button policy
• Design system

Tech Stack

• Node.js 22
• Fastify
• SQLite via better-sqlite3
• TypeScript
• Vanilla HTML/CSS/JavaScript
• Lucide-style SVG icons

License

MIT