
vibekanban
Orchestrate AI coding with Vibe Kanban: tasks, review, sessions, workspaces, and isolated git worktrees. Use when managing AI-generated code in isolated environments, planning coding tasks, reviewing AI output, or configuring Vibe Kanban workspaces and agents. Keywords: Vibe Kanban, AI orchestration, worktrees.
- Installs
- 41
- Last 24h
- —
- Trending rank
- —
- Security
- Not audited
Install
Works with every agent the skills CLI supports.
npx skills add https://github.com/itechmeat/llm-code --skill vibekanban- 01
Install
Run the command in your project. The CLI asks which agents to install for.
- 02
Describe the task
Ask your agent for something this skill covers. It loads vibekanban automatically when relevant.
- 03
Review the result
The agent follows the SKILL.md steps, scripts and references bundled with the skill.
SKILL.md
The instructions your agent reads when the skill is triggered.
Vibe Kanban
Orchestration platform for AI coding agents. Plan, review, and manage AI-generated code in isolated git worktrees.
Links
Quick Navigation
| Topic | Reference |
|---|---|
| Installation & Setup | getting-started.md |
| Workspaces (Beta) | workspaces.md |
| Projects, Tasks, Review | core-features.md |
| Subtasks, Attempts, Conflicts | advanced-features.md |
| Settings, Agents, Tags | configuration.md |
| GitHub, Azure, VSCode, MCP | integrations.md |
| Common Issues | troubleshooting.md |
| Best Practices | vibe-guide.md |
Quick Start
Install & Run
npx vibe-kanban
npx vibe-kanban remains the canonical cross-platform launch path. As of v0.1.30, the packaged desktop app is moving onto a Tauri v2 shell with built-in auto-update behavior for installed desktop builds.
Release Highlights (0.1.15 → 0.1.28)
- MCP server expanded: richer issue retrieval (
get_issueincludes tags/relationships/sub-issues) and new workspace/relationship/tag tools (e.g.update_workspace, relationship + tag mutations). - Frontend routing migrated to TanStack Router (navigation/URLs may differ from older screenshots).
- Reliability improvements around Electric fallback and cancellation/tab-switch handling.
- Review/diff UI improvements (annotation width fixes for horizontal scrolling).
- Remote Access: pair a host via a code and access its workspaces from another device.
- Workspace actions became safer around spin-off/duplicate flows by preserving branch and executor configuration more reliably.
- Active agent runs can now accept image attachments, which is useful when feeding UI state, screenshots, or design feedback back into an in-progress session.
- Transport reliability improvements: better relay disconnect handling and signing-session refresh.
- Claude: bumped the default Sonnet model to 4.6.
- Worktrees: preserve worktree path across cleanup to maintain Claude Code session continuity.
- Workspaces: improved workspace logs capture (root execution-process provider).
- Mobile UI fixes and Remote Access docs refresh.
Release Highlights (0.1.44)
- Claude Code CLI pin moved forward again and the Codex executor surface was refreshed around Codex
0.124.0. - Legacy project routes are being sunset to an export-only page, so deep links or screenshots from older project pages may no longer match the current UX.
- Execution logs are pre-registered before normalized streaming begins, which improves log availability during live runs.
- Model selector ordering is now alphabetical, which changes some small but visible operator workflows in setup/review dialogs.
Release Highlights (0.1.37 → 0.1.43)
- Executor surface refreshed with newer pinned versions for Codex, OpenCode, and Claude Code.
- Claude Code adds an
xhigh effortvariant. - Cross-compile/runtime fixes improve Windows/MSVC and musl build reliability.
Release Highlights (0.1.31 → 0.1.32)
- Desktop/Tauri: Windows desktop app support lands, Tauri notifications are available, and desktop zoom behavior is reworked around font-size scaling.
- Workspaces: sessions can be renamed and auto-named; draft workspace-creation preference is persisted.
- Attachments: arbitrary attachments are supported across workspaces and issues.
- MCP: issue filters are available through the MCP surface.
- Remote/admin: remote audit logging is added, short issue IDs become org-scoped, and analytics are disabled when PostHog config is blank.
- Editor UX: WYSIWYG typeahead, toolbar, and issue-description editing flows are improved.
Opens browser automatically. Use PORT=8080 npx vibe-kanban for fixed port.
Prerequisites
- Node.js LTS
- Authenticated coding agent (Claude Code, Codex, Gemini, etc.)
- GitHub CLI (
gh) for PR integration
First Steps
- Authenticate with a coding agent externally
- Run
npx vibe-kanban - Complete setup dialogs
- Create project from existing git repo
- Add tasks and start executing
Two UI Modes
Classic Kanban (Tasks)
Traditional board with columns: To do → In Progress → In Review → Done
Updates in v0.1.7:
- Per-project Kanban views and a refreshed filter dialog
- Sub-issues and Workspaces visibility toggles moved to the filter bar
Workspaces (Beta) — NEW
Modern interface with:
- Sessions: Multiple conversation threads per workspace
- Command Bar:
Cmd/Ctrl + Kfor all actions - Workspace Notes: Document requirements and decisions
- Multi-repo support: Work across multiple repositories
- Integrated Terminal: PTY-backed terminal with shell support
- Session Dropdown: Agent icons displayed next to session titles
- Sidebar filters: Filter by project and PR status (including “No project”)
- Needs Attention: Includes workspaces with unseen activity
Switch between modes via Command Bar → "Open in Old UI"
Core Concepts
Git Worktrees
Each task/workspace runs in an isolated git worktree:
- Agents can't interfere with each other
- Safe from main branch changes
- Automatic cleanup after completion
Task Flow (Classic)
To do → In Progress → In Review → Done
- To do: Task created
- In Progress: Agent executing
- In Review: Agent finished, awaiting review
- Done: Merged or PR merged
Task Attempts
One task can have multiple attempts:
- Different agent
- Different branch
- Fresh conversation context
Supported Agents
| Agent | Variants |
|---|---|
| Claude Code | DEFAULT, PLAN, ROUTER, XHIGH |
| Codex | DEFAULT, HIGH |
| Gemini | DEFAULT, FLASH |
| GitHub Copilot | DEFAULT |
| Amp | DEFAULT |
| Cursor Agent | DEFAULT |
| OpenCode | DEFAULT |
| Qwen Code | DEFAULT |
| Droid | DEFAULT |
| Antigravity | DEFAULT |
Select agent when creating task attempt or workspace session.
Project Configuration
Setup Scripts
Run before agent execution (e.g., npm install, cargo build).
Dev Server Scripts
Start dev server for preview (e.g., npm run dev).
Cleanup Scripts
Run after agent finishes (e.g., npm run format).
Copy Files
Files to copy from main project to worktree (e.g., .env).
Task Creation
Press C or click + to create task
Options:
- Create Task: Add to board only
- Create & Start: Add and immediately execute with default agent
Task Tags
Reusable snippets via @mention:
- Type
@in description - Select tag from dropdown
- Content inserted at cursor
Task description editor supports markdown paste and preserves inline code formatting.
Code Review
- Task moves to "In Review" when agent finishes
- Click Diff icon to view changes
- Click
+on any line to add comment - Submit all comments together
- Task returns to "In Progress" for fixes
Git Operations
| Action | Description |
|---|---|
| Merge | Merge to target branch |
| Create PR | Open PR on GitHub/Azure |
| Rebase | Update with target branch changes |
| Push | Push additional changes to PR |
Preview Mode
Test web apps without leaving Vibe Kanban:
- Configure dev server script in project settings
- Click "Start Dev Server" in Preview tab
- View app in embedded iframe
- Install
vibe-kanban-web-companionfor component selection
Keyboard Shortcuts
| Key | Action |
|---|---|
C | Create task |
⌘/Ctrl + Enter | Submit/Send message |
k/j | Navigate up/down in column |
h/l | Navigate left/right between columns |
Enter | Open task |
⌘/Ctrl + S | Focus search |
MCP Integration
Add MCP Servers to Agents
Settings → MCP Servers → Select agent → Add servers
Vibe Kanban MCP Server
Expose Vibe Kanban to external MCP clients:
{
"mcpServers": {
"vibe_kanban": {
"command": "npx",
"args": ["-y", "vibe-kanban@latest", "--mcp"]
}
}
}
MCP tools include: list_workspaces, update_workspace, list_projects, list_issues, get_issue, update_issue, plus tag and relationship helpers.
Critical Safety Note
Vibe Kanban runs agents with --dangerously-skip-permissions/--yolo by default for autonomous operation. Each task runs in isolated worktree, but agents can still perform system-level actions. Review work and keep backups.
Critical Prohibitions
- Do not skip agent authentication before first use
- Do not ignore worktree isolation benefits
- Do not forget to configure setup/cleanup scripts for dependencies
- Do not mix multiple agents on same task without new attempts
- Do not ignore rebase conflicts — resolve or abort
Links
FAQ
What does the vibekanban skill do?
Orchestrate AI coding with Vibe Kanban: tasks, review, sessions, workspaces, and isolated git worktrees. Use when managing AI-generated code in isolated environments, planning coding tasks, reviewing AI output, or configuring Vibe Kanban workspaces and agents. Keywords: Vibe Kanban, AI orchestration, worktrees.
How do I install vibekanban?
Run npx skills add https://github.com/itechmeat/llm-code --skill vibekanban in your project. The skills CLI detects the agents you use and copies the skill into the right folder. For Claude Code you can also place the skill folder at ~/.claude/skills/vibekanban/ (personal) or .claude/skills/vibekanban/ (project).
Which AI agents can use vibekanban?
Any agent that supports the open Agent Skills format (a folder with a SKILL.md file), including Claude Code, Codex, Cursor, GitHub Copilot, Gemini CLI, Windsurf, Cline, OpenCode and more.
Is vibekanban safe to install?
No partner security audit is available yet. Skills can run commands and scripts, so read the SKILL.md and any bundled files before installing.
Who maintains vibekanban?
It is published by itechmeat in itechmeat/llm-code. It has 41 recorded installs.
More from itechmeat/llm-code
react-testing-library
itechmeat/llm-code
React Testing Library: user-centric component testing with queries, user-event simulation, async utilities, and accessibility-first API. Use when writing React component tests, selecting elements by role/label/text, simulating user events, or testing async UI behavior. Keywords: React Testing Library, @testing-library/react, user-event, queries, render.
openspec
itechmeat/llm-code
OpenSpec artifact-driven workflow. Covers OPSX commands, schemas, project config. Use when applying the artifact-driven workflow (OPSX), planning or reviewing changes based on artifact dependencies, or working with OPSX commands and schema/template resolution. Keywords: OPSX, artifact graph, /opsx:.
mantine-dev
itechmeat/llm-code
Mantine UI library for React: 100+ components, hooks, forms, theming, dark mode, CSS modules, and Vite/TypeScript setup. Use when building React applications with Mantine components, configuring theming/dark mode, or working with Mantine hooks and forms. Keywords: Mantine, React, UI components, CSS modules, theming.
social-writer
itechmeat/llm-code
Open the skill to read its SKILL.md instructions.
base-ui
itechmeat/llm-code
Base UI unstyled React components. Covers forms, menus, overlays, composition. Use when building accessible, unstyled React UI components with @base-ui/react, composing with render props, or implementing custom-styled form controls, menus, and overlays. Keywords: @base-ui/react, render props, unstyled components.
refine-dev
itechmeat/llm-code
Refine.dev headless React framework for CRUD apps: data providers, resources, routing, authentication, hooks, and forms. Use when building admin panels, dashboards, or internal tools with React and various backends (REST, GraphQL, Supabase, Strapi), or configuring Refine data providers and authentication. Keywords: Refine, CRUD, admin panel, data provider, React.
Popular code quality & workflow skills
grill-me
mattpocock/skills
A relentless interview to sharpen a plan or design.
improve-codebase-architecture
mattpocock/skills
Scan a codebase for deepening opportunities, present them as a visual HTML report, then grill through whichever one you pick.
handoff
mattpocock/skills
Compact the current conversation into a handoff document for another agent to pick up.
domain-modeling
mattpocock/skills
Build and sharpen a project's domain model. Use when discussing codebase terminology, writing or editing a GLOSSARY.md, or recording or editing an ADR.
codebase-design
mattpocock/skills
Shared vocabulary for designing deep modules. Use when the user wants to design or improve a module's interface, find deepening opportunities, decide where a seam goes, make code more testable or AI-navigable, or when another skill needs the deep-module vocabulary.
diagnosing-bugs
mattpocock/skills
Diagnosis loop for hard bugs and performance regressions. Use when the user says "diagnose"/"debug this", or reports something broken/throwing/failing/slow.