Upcoming Event

Ultimate Agents Hackathon

Sat, Oct 25 • 9 AM–9 PM ET•Betaworks NYC•$5k+ prizes

Move faster with
Claude Code
Move faster withClaude Code

Vibe Kanban lets you run coding agents in parallel without conflicts, and perform code review through our diff tool. Now you can focus on planning and quality instead of watching terminal logs.

Install Node.js 18+ then run:

$npx vibe-kanban

Vibe kanban is the biggest increase I've had in productivity since cursor.

Luke Harries

Growth Lead at Eleven Labs

Choose Your Coding Agent

Works seamlessly with all your favorite AI coding agents.

Run local coding agents in parallel

Instead of staring at coding agent logs you can focus on planning and reviewing work while your coding agents work in the background.

Give agents their own workspace

Agents work in separate branches and workspaces on your machine (powered by git worktrees), so they never step on each other's toes during parallel execution. Safely merge changes back when tasks complete.

Review agent changes like human code

Built-in diffs let you inspect, edit, and approve agent work just like any pull request.

Stay ahead of the curve

Thousands of top developers use Vibe Kanban daily to orchestrate AI coding agents

HenryAllen04/Sigma-Saunas

improved dashboard (vibe-kanban)

britannio

Contributor

merged

HenryAllen04/Sigma-Saunas

unify the voice page with the root page (vibe-kanban)

britannio

Contributor

merged

bankjaneo/marker

Implement unload llama.cpp model to free up vram before OCR (vibe-kanban)

We just finished implement FREE_VRAM_ON_IDLE=true on latest commit. Now to improve vram usage further on system with not much vram. We should implement unload model for user using llama-swap (https://github.com/mostlygeek/llama-swap). There is already an endpoint for that at /models/unload Add new environment variable (OPENAI_UNLOAD_MODEL=true) The process should be 1. Unload llama-swap model (if OPENAI_UNLOAD_MODEL=true) 2. Load up marker OCR model (if FREE_VRAM_ON_IDLE=true) 3. Do an OCR 4. Free up marker OCR model (if FREE_VRAM_ON_IDLE=true) So what we should do here is about Step 1 only. What do you think?

bankjaneo

Contributor

merged

bankjaneo/marker

Improve VRAM usage (vibe-kanban)

The current marker_server on docker gpu-enabled version will always occupied VRAM even on idle. Is there a way to let marker load model to vram when doing the task and unload it when finish? This will be useful for machine with not much VRAM. Maybe user can control this behavior by setting environment variable - FREE_VRAM_ON_IDLE=true

bankjaneo

Contributor

merged

hoanglt1223/vn-real-estate-analyze

complete todo.md (vibe-kanban)

hoanglt1223

Contributor

merged

HenryAllen04/Sigma-Saunas

improved dashboard (vibe-kanban)

britannio

Contributor

merged

HenryAllen04/Sigma-Saunas

unify the voice page with the root page (vibe-kanban)

britannio

Contributor

merged

bankjaneo/marker

Implement unload llama.cpp model to free up vram before OCR (vibe-kanban)

bankjaneo

Contributor

merged

bankjaneo/marker

Improve VRAM usage (vibe-kanban)

bankjaneo

Contributor

merged

hoanglt1223/vn-real-estate-analyze

complete todo.md (vibe-kanban)

hoanglt1223

Contributor

merged

IAFahim/Ludo

Add client and server example. for Rest, SingnalR, Websoket. Each on serparte Soultion (vibe-kanban)

IAFahim

Contributor

merged

FlexNetOS/noa_ark_os

fix(vibe-kanban): repair imports and dependencies for build

## Summary - add ui/shared to the pnpm workspace and supply type shims for shared react and mitt usage - repair broken relative imports in the vibe-kanban API routes and export the session state type - localize the AI controller and router modules, tighten rate limiter and sqlite helpers, and wire mitt typings into the app config ## Testing - pnpm --filter vibe-kanban build ------ [Codex Task](https://chatgpt.com/codex/tasks/task_e_69147f93766083219b1fbb7f662b7ece)

FlexNetOS

Contributor

merged

RSOLV-dev/rsolv-action

bump to maintained node lts (vibe-kanban)

node 18 is unmaintained; ensure we use more recent everywhere including ci. tdd.

arubis

Contributor

merged

bassa01/google-cloud-mcp

fix error (vibe-kanban)

bassa01

Contributor

merged

bassa01/SpannerSpy

Add Test (vibe-kanban)

bassa01

Contributor

merged

IAFahim/Ludo

Add client and server example. for Rest, SingnalR, Websoket. Each on serparte Soultion (vibe-kanban)

IAFahim

Contributor

merged

FlexNetOS/noa_ark_os

fix(vibe-kanban): repair imports and dependencies for build

FlexNetOS

Contributor

merged

RSOLV-dev/rsolv-action

bump to maintained node lts (vibe-kanban)

node 18 is unmaintained; ensure we use more recent everywhere including ci. tdd.

arubis

Contributor

merged

bassa01/google-cloud-mcp

fix error (vibe-kanban)

bassa01

Contributor

merged

bassa01/SpannerSpy

Add Test (vibe-kanban)

bassa01

Contributor

merged

RSOLV-dev/rsolv-action

v3.8.0: Fix issue grouping to prioritize CRITICAL over HIGH severity (vibe-kanban)

When max_issues=1, grouping chose 3 HIGH over 1 CRITICAL. Should prioritize by severity: CRITICAL > HIGH > MEDIUM > LOW

arubis

Contributor

merged

Arbust908/the-mystaran

Save found links into a .json file (vibe-kanban)

Edit the `crawler` utils to build an enpoint that will get all links from the alexandrin page and save it to a `found_links.json` file

Arbust908

Contributor

merged

gharam1234/14th_homework

t (vibe-kanban)

/Users/gharam12/Desktop/my-homework/14th_homework/monorepo/@commons/ui/src/commons/components/textarea/prompts/prompt.101.ui.txt

gharam1234

Contributor

merged

RSOLV-dev/rsolv-action

v3.8.0: Fix issue grouping to prioritize CRITICAL over HIGH severity (vibe-kanban)

When max_issues=1, grouping chose 3 HIGH over 1 CRITICAL. Should prioritize by severity: CRITICAL > HIGH > MEDIUM > LOW

arubis

Contributor

merged

Arbust908/the-mystaran

Save found links into a .json file (vibe-kanban)

Edit the `crawler` utils to build an enpoint that will get all links from the alexandrin page and save it to a `found_links.json` file

Arbust908

Contributor

merged

gharam1234/14th_homework

t (vibe-kanban)

/Users/gharam12/Desktop/my-homework/14th_homework/monorepo/@commons/ui/src/commons/components/textarea/prompts/prompt.101.ui.txt

gharam1234

Contributor

merged

Create tasks from other apps

Connect your favorite MCP client like Claude Desktop to access Vibe Kanban's capabilities directly from your personal AI assistant and extend your workflow beyond the interface.

This project is underrated. I've found it to be useful and fun.

Hamel Husain

Leading AI Consultant, 20 Years of Experience

Work from your IDE

Monitor agents, interact with them, perform code review, and see running processes directly in your IDE.

Motivation

Coding agents are reliable enough to be the default: We've built and tested coding agents for years and realised recently that reliability crossed an invisible threshold that means we now prefer starting most tasks with coding agents.

Coding agents are going to get much better: Coding agents have improved rapidly, and we expect this trend to continue. Imagine that in six months, 50% of the current failure modes of coding agents will get fixed, and six months after that another 50%. What will we (engineers) be spending our time on in that world? What tools would help us do that work most efficiently?

The SOTA will change: Foundational model labs and application layer startups will continuously leap-frog each other to build state-of-the-art coding agents over the next 12m+. You shouldn't have to change your workflow or behaviour to be able to try out different coding agents.

FAQs

Common questions about Vibe Kanban and how to get the most out of your AI coding agents

Yes, Vibe Kanban runs locally on your machine and doesn't send your code to external servers.

Vibe Kanban supports a wide range of coding agents including Claude Code, Gemini CLI, Cursor CLI, Amp, OpenAI Codex, Qwen Code, Opencode, and GitHub Copilot. You can find the complete list of supported agents in our documentation at vibekanban.com/docs. You can use multiple agents simultaneously and switch between them for different tasks.

Vibe Kanban itself is free and open source. You only pay for the underlying AI services (like Claude Code, OpenAI Codex, etc.) that you choose to use with your agents. There are no additional fees or subscriptions for using Vibe Kanban.

Absolutely! One of Vibe Kanban's key features is parallel execution. You can run multiple coding tasks simultaneously across different agents and projects. Each task runs in its own git worktree on your machine.

The VS Code extension brings the task view into your IDE. Search for 'Vibe Kanban' in the VS Code marketplace and it will automatically connect to your local Vibe Kanban instance.

MCP (Model Context Protocol) works with Vibe Kanban in two ways: First, you can connect MCP servers to your coding agents so they can access external resources and tools. Second, Vibe Kanban itself exposes an MCP server that you can connect to from third-party MCP clients like Claude Desktop, allowing you to seamlessly create tasks from third-party clients.

Yes! Vibe Kanban allows you to create reusable task templates to standardise your development processes. You can define templates for common tasks like bug fixes, feature implementations, or refactoring, which helps boost team productivity and consistency.

Yes, Vibe Kanban has seamless GitHub integration. It can automatically create pull requests after task completion, handle rebases and merges with one-click operations, and maintain clean git history. This makes it easy to incorporate AI-assisted development into your existing workflow.

Vibe Kanban provides real-time monitoring and control over agent execution. You can pause, resume, or stop agents at any time. The system also provides detailed logs and step-by-step visibility, making it easy to understand what went wrong and intervene when necessary.

Ready to get started?

Start managing your AI coding agents with Vibe Kanban.

Install Node.js 18+ then run:

$npx vibe-kanban

Ultimate Agents Hackathon

Move faster with
Claude Code
Move faster withClaude Code

Luke Harries

Choose Your Coding Agent

Run local coding agents in parallel

Give agents their own workspace

Review agent changes like human code

Stay ahead of the curve

Create tasks from other apps

Hamel Husain

Work from your IDE

Motivation

FAQs

Is Vibe Kanban secure?

Which coding agents does it support?

How much does it cost?

Can I use multiple agents at the same time?

How does the VS Code extension work?

What is MCP and how does it work with Vibe Kanban?

Can I customize task templates?

Does it integrate with GitHub?

What if my coding agent fails or gets stuck?

Ready to get started?

Learn More

Ultimate Agents Hackathon

Move faster withClaude CodeClaude CodeMove faster withClaude CodeClaude Code

Luke Harries

Choose Your Coding Agent

Run local coding agents in parallel

Give agents their own workspace

Review agent changes like human code

Stay ahead of the curve

Create tasks from other apps

Hamel Husain

Work from your IDE

Motivation

FAQs

Is Vibe Kanban secure?

Which coding agents does it support?

How much does it cost?

Can I use multiple agents at the same time?

How does the VS Code extension work?

What is MCP and how does it work with Vibe Kanban?

Can I customize task templates?

Does it integrate with GitHub?

What if my coding agent fails or gets stuck?

Ready to get started?

Learn More

Move faster with
Claude Code
Move faster withClaude Code