Show HN: Vibe Kanban – Kanban board to manage your AI coding agents

Show HN: Vibe Kanban – Kanban board to manage your AI coding agents(github.com)

195 points by louiskw 325 days ago | 132 comments

Hey HN! I'm Louis, one of the creators of Vibe Kanban.

We started working on this a few weeks ago. Personally, I was feeling pretty useless working synchronously with coding agents. The 2-5 minutes that they take to complete their work often led me to distraction and doomscrolling.

But there's plenty of productive work that we (human engineers) could be doing in that time, especially if we run coding agents in the background and parallelise them.

Vibe Kanban lets you effortlessly spin up multiple coding agents. While some agents handle tasks in the background, you can focus on planning future work or reviewing completed tasks.

After a few weeks of internal dog fooding and sharing it with friends, we've now open-sourced Vibe Kanban, and it's stable enough for day-to-day use.

I'd love to hear your feedback, feel free to open an issue on the github and we'll respond ASAP.

gpm 325 days ago |

Hmm, analytics appear to default to enabled: https://github.com/BloopAI/vibe-kanban/blob/609f9c4f9e989b59...

It is harvesting email addresses and github usernames: https://github.com/BloopAI/vibe-kanban/blob/609f9c4f9e989b59...

Then it seems to track every time you start/finish/merge/attempt a task, and every time you run a dev server. Including what executors you are using (I think this means "claude code" or the like), whether attempts succeeded or not and their exit codes, and various booleans like whether or not a project is an existing one, or whether or not you've set up scripts to run with it.

This really strikes me as something that should be, must legally be in many jurisdictions, opt in.

louiskw 324 days ago | |

That's fair feedback, I have a PR with a very clear opt-in here https://github.com/BloopAI/vibe-kanban/pull/146

I will leave this open for comments for the next hour and then merge.

TeMPOraL 324 days ago | | |

Nice, I vote for merging it :).

It really doesn't hurt to be honest about this and ask up-front. This is clear enough and benign enough that I'd actually be happy to opt-in.

smcleod 324 days ago | | |

Good on you for taking action on this kind of feedback!

arresin 325 days ago | |

Thanks, really appreciate the heads up. I put devs who do this on a personal black list for life.

I think also that this would be better as an mcp tool / resource. Let the model operate and query it as needed.

willsmith72 324 days ago | | |

It's the email/username harvesting that you mean right? Or do people also have something against anonymised product analytics?

swyx 324 days ago | |

could you point me to what jurisdictions require analytics opt in esp for open source devtools? thats not actually something ive seen as a legal requirement, more a community preference.

eg ok we all know about EU website cookie banners, but i am more ignorant about devtools/clis sending back telemetry. any actual laws cited here would update me significatnly

47282847 323 days ago | | |

GDPR is not about cookies but about privacy in general. It’s an easy read, and yes, it applies to software and telemetry as much as it applies to websites and cookies, and it applies to anyone providing services and tools to Europeans.

"Personal data is information that relates to an identified or identifiable individual. If you cannot directly identify an individual from that information, then you need to consider whether the individual is still identifiable. You should take into account the information you are processing together with all the means reasonably likely to be used by either you or any other person to identify that individual."

gpm 324 days ago | | |

I mean, you've labelled one big one already with the GDPR covering a significant fraction of the world - and unlike your average analytics "username and email address" sounds unquestionably identifying/personal information.

Where I live I think this would violate PIPEDA, the Canadian privacy law that covers all business that do business in any Canadian province/territory other than BC/Alberta/Quebec (which all have similar laws).

There's generally no exception in these for "open source devtools" - laws are typically still laws even if release something for free. The Canadian version (though I don't think the GDPR does) has an exception for entirely non-commercial organizations, but Bloop AI appears to be a commercial organization so it wouldn't apply. It also contains an exception for business contact information - but as I understand it that is not interpreted broadly enough to cover random developers email addresses just because they happen to be used for a potentially personal github account.

Disclaimer: Not a lawyer. You should probably consult a lawyer in the relevant jurisdiction (i.e. all of them) if it actually matters to you.

jjangkke 324 days ago | |

analytics stuff is fine but the email harvesting/github username appears to be illegal especially if its done without notifying the user?

great catch, many open source projects appear to be just an elaborate lead gen tool these days.

janoelze 324 days ago | |

fork, task claude to remove all github dependence, build.

gpm 324 days ago | | |

I did this locally to try it out :) Also stubbed out the telemetry and added jj support. "Personalizing" software like this is definitely one of LLMs superpowers.

I'm not particularly inclined to publish it because I don't want to associate myself with a project harvesting emails like this.

hsbauauvhabzb 324 days ago | | |

Use a telemetry backed tool to remove telemetry from another telemetry backed tool?

swalsh 325 days ago |

I built something similar for my own workflow. Works okay. The hard part is as you scale, you end up with compounded false affirmatives. Model adds some fallback mechanism that makes it work, tests pass, etc. The nice part is you can ask models to review the code from others, call out fallbacks, hard coding, stuff like that. It does a good job at identifying buried bodies. But if you dig up a buried body, I'd manually confirm it was properly disposed of as the models usually hid the body in the first place because they needed some input they didn't have, got confused or ran into an issue.

oc1 325 days ago |

We need something like a kitchen brigade in software - one who writes the vibe code tickets (Chef de Vibe), one who reviews the vibe code (Sous-Vibe), one who oversees the agents and restarts them if they get hung up (Agent de Station). We could theoretically smash thousand tickets a day with this principle

ggordonhall 325 days ago | |

Completely agree!

You can actually use a coding agent to create tickets from within Vibe Kanban. Add the Vibe Kanban MCP server (from MCP settings) and ask the agent to plan a task and write tickets.

atavistically 324 days ago | |

c.f. "Surgical Team" in 'The Mythical Man-Month' by Fred Brooks. That book is perennially relevant.

lharries 325 days ago |

I used this last week and it's excellent - feels like the same increase in productivity increase from when I first used Cursor.

Are you thinking of doing a hosted version so I can have my team collab on it?

And I found I could open lots of PRs at once but they often need to be dependent on each other - and then I want to make a change to the first one. How are you thinking of better managing that flow?

louiskw 325 days ago | |

Yeah I think giving the option to move execution to the cloud makes a lot of sense, I already find my macbook slowing down after 4 concurrent runs, mainly rustc.

Also now we're pushing many more PRs think we defo need better ways to stack and review work. Will look into this asap

hddbbdbfnfdk 325 days ago | |

Very productive increase sirs! Whole team well promoted.

adastra22 324 days ago |

> AI coding agents are increasingly writing the world's code and human engineers now spend the majority of their time planning, reviewing, and orchestrating tasks.

Is this really the case?

sexeriy237 321 days ago | |

No, if we can review 10 PRs a day and AI writes one of them, we now have to review 11 PRs

barbazoo 325 days ago |

> human engineers now spend the majority of their time planning, reviewing, and orchestrating tasks

This feel like much too broad a statement to be true.

uxamanda 325 days ago |

If you use gitlab, you can use the command line "glab" tool to have agents work from the built in kanban. They can open and close tasks, start MRs off of them etc. It's not as integrated as this tool, but works well with a mix of humans and robots.

louiskw 324 days ago | |

Interesting, hadn't heard of that. Would better GitLab support be useful in Vibe Kanban?

PaulIH 322 days ago | | |

Yes, being able to use Gitlab as a provider would mean that we would jump on the tool, being Gitlab-based. :)

deepdarkforest 325 days ago |

This is a launch by a YC company that converts enterprise cobol code into java. Maybe it's my fault, but i tried every single coding agent with a variety of similar tools and whenever i try to parallelize, they clash while editing files simultaneously, i lose mental context of what's going on, they rewrite tests etc.

It's chaos. Thats fine if you are vibe coding an unimportant nextjs/vercel demo, but i'm really sceptical of all this stance that you should be proud of how abstracted you are from code. A kanban board to just shoot off as many tasks as possible and just quickly read over the PR's is crazy to me. If you want to appear a serious company that should be allowed to write enterprise code, imo this path is so risky. I see this in quite a few podcasts, tweets etc. People bragging how abstracted they are from their own product anymore. Again, maybe i am missing something, but all of this github copilot/just reviewing like 10 coding agents PR's just introduces so much noise and slop. Is it really what you want your image to be as a code company?

doritosfan84 325 days ago |

The permissions this asks for feel kinda insane to me. Why does a kanban board need to see the code or my deploy keys among other things?

jeltz 325 days ago | |

I would assume because it was vibe coded.

gpm 325 days ago | | |

More generously I'd assume because

- It's an early prototype so they haven't dealt with fine grained permissions

- They really do want to do things like access private repos with it themselves

- They really do want the ability to do things like checkout code, create PRs, etc... and that involves a lot of permission.

TeMPOraL 324 days ago | |

Because it's not "a kanban board"? It's a coding agent orchestrator that's made in the shape of a Kanban board.

You might be right that this app asks for excessively broad privileges, but your case would be much stronger if it wasn't backed by an absurdly disingenuous argument.

Shypangz 315 days ago |

Waiting around for AI agents to finish can kill productivity. You might want to try leads app to organize tasks and keep track of progress smoothly. It helped me stay focused while letting background work run without distractions.

_jayhack_ 325 days ago |

Very cool and interesting project. Ideas like this are a threat to traditionally-conceived project management platforms like Linear; that being said, Linear and others (Monday, ClickUp, etc.) are pushing aggressively into UX built for human/AI collaboration. I guess the question is how quickly they can execute and how many novel features are required to properly bring AI into the human project workspace

louiskw 325 days ago | |

Cheers! Smaller teams, more infrastructure, more testing, tasks requiring review in minutes not days - the features are just totally different for the new world than what legacy PM tools are optimised for, and who they have to continue to serve.

skeeter2020 324 days ago |

If I multiply my 100x productivity gains from using AI with your 10x increase what am I supposed to do with all that free time?

ffsm8 324 days ago | |

Maybe Tony can inspire you?

( ◠ ‿ ・ ) —

https://youtube.com/shorts/YBAcvRV7VSM?si=jp2hZvFIVo-vSdu6

Vaslo 324 days ago | | |

Lol - nice find

akifq 322 days ago |

Ran Vibe Kanban on a side repo over the weekend it’s slick, but I tripped on one thing: all the agents write into the same checkout, so parallel cards can clobber each other when they touch the same files. I’ve been spinning up a quick `git worktree` per card as a stop gap.

Outside of that, really enjoying the flow.

BTW, if anyone’s juggling multiple agents and wants a “quarantine” layer before changes hit Git, I’m hacking on SAV (Shadow VCS)open source, risk scores every commit and keeps them sandboxed until you approve. Happy to swap notes. https://github.com/AKIFQ/sav

BoiledCabbage 322 days ago | |

How is the isolation layer of a shadow vcs different/better than just checking into a feature branch and not pushing?

Not saying there isn't value in the risk assessment part, but, I'm asking specifically about the change isolation.

louiskw 322 days ago | |

Are you sure you're using vibe-kanban? The project spins up a worktree for each task attempt specifically to solve this problem.

jackbridger 324 days ago |

This is interesting. I’m using Claude Code a fair bit and have found writing specs to be more effective than promoting and this feels closer to that. I can see the appeal of this for simple tasks now and maybe increasingly bigger tasks as models get better.

louiskw 324 days ago | |

Very much a bet that things are going to get much much better very quickly

randysalami 325 days ago |

I tried to build something similar but in a peer-to-peer fashion and for humans + AI. It was supposed to be like a Kanban board that could scale to any team size and use Planning AI to ingest/match/monitor work realtime across teams and agents. I ran out of steam and couldn’t get funding but here is the prototype version:

https://postwork-alpha.vercel.app/

User: maryann.biaggioli@astarconsulting.com

Pass: Test1234!

I never got to a point where I actually integrated AI agents (weren’t as good at the time) but it’s cool to see it working in the real world!

koistya 321 days ago |

Cool project! Been messing with AI agents for code stuff, and this Kanban setup for parallel tasks looks super handy—gonna try it on my next hack. Quick Q: handling token tracking per user/org? Got tons of experience from building UsagePilot (open-core lib for AI metrics). Happy to chat if you need a second pair of hands on that front!

diggan 325 days ago |

That's an interesting idea and looks neat! I have my own developed agent running locally in containers, and currently use GitHub issues+pull requests for coordinating all the asynchronous work. Do you have any pointers on the approach I should take if I basically have something like a service already running for this, and I just want to hook up your UI to use it instead? Just some broad pointers on what would be required would be most helpful already!

slig 325 days ago |

How does this compare with Backlog.md? [1]

[1]: https://news.ycombinator.com/item?id=44483530

ggordonhall 325 days ago | |

Hi, co-author of this project here!

In Vibe Kanban you can directly interact with Claude Code from within the Kanban board. E.g. you can write out a ticket, hit a button to run it locally with Claude Code/Gemini etc., watch its responses, and then review any diffs that it generated.

slig 325 days ago | | |

Thank you! I was taking a look at the docs, and I'm going to play with it later today. Thanks for sharing and congrats on shipping!

mrlesk 321 days ago | |

Hi. I’m the author of backlog.md . Your team did a great job on Vibe Kanban. It looks great and has a cool name. I like to see more tools trying to solve the human-ai agent interaction for building projects.

I wish you all the best with this tool!

helsinki 324 days ago |

So far, it’s relatively bug-free well-written code that I’ve forked to work behind the walls of a hedge fund, and it works, but the reality is that it doesn’t provide anything that some terminal windows and git worktrees can’t offer. Am I missing something?

You really need to add more features, because I struggle to find a compelling reason for advanced users to use it.

notarobot123 324 days ago |

I'd be interested in how it _feels_ to run with this workflow for an extended period.

If you're doing code reviews, writing new tickets, evaluating progress and guiding the overall structure of a project, do you loose something important or is it genuinely a satisfying way of working which you could imagine sustaining for the long-term?

mahsima 324 days ago |

Great

Add it to this lists, i think it helps:

https://github.com/tokyo-dal/awesome-ai-coding-tools

https://github.com/devtoolsd/awesome-devtools

and any awesome lists related to ai development

gpm 325 days ago |

I definitely don't feel like the models are reliable enough that I'd be more productive running them in parallel like this yet, but I can see a future where I want this.

Their reliability probably varies a lot depending on what you are using them for - so maybe I'm just using them in more difficult (for claude) domains.

louiskw 324 days ago | |

Yes I generally cherry pick the easier 50% of my backlog and work on those with Vibe Kanban, and the other 50% is still manual or happens with coding agent but with a human-in-the-loop.

This is a bet that coding agents will continue to get better, and this feels like the right time to try and figure out the interface.

sqs 324 days ago |

This is really cool. I used Vibe Kanban with Amp to update some of our docs and UI components, and it was great.

louiskw 324 days ago | |

I would say conservatively that 80% of Vibe Kanban has been built by Amp

FailMore 325 days ago |

Do you think you will keep it free or can you see a business model developing around it? If so, what do you think it would be? / How would you split paid tiers vs free users? Not a big deal to me...!! But I'm curious how one might commercialise these types of free/open source projects

louiskw 324 days ago | |

I could see there being a long term free offering that doesn't cost us compute or tokens, and probably some other offerings that actually do use resources and would make sense to build a business around.

But that's not a today problem, we just want to absorb feedback and iterate until we build the ultimate tool for working with these coding agents.

swyx 325 days ago |

Louis did a long form chat and vibe coded extra features in our discord meetup: https://www.youtube.com/watch?v=NCksand7Iwo for more info on this

louiskw 325 days ago | |

"11 days ago" this feels like a year ago :O

peadarohaodha 325 days ago |

Would love to have a good selection of keyboard shortcuts! Power Vibe Kanban

louiskw 325 days ago | |

Good point - there are a few random ones already but I will ask a coding agent to implement this comprehensively

sails 324 days ago |

If you could consider how a team could use linear in conjunction with this that would be very interesting.

I would imagine matching tickets/conversations from Linear mcp to use as an overlay of context

remram 325 days ago |

Why do you need to "manage" your coding agents like they are people? How long does it take them to do a task once prompted, in the background?

Don't you just prompt an immediately review the result?

scotty79 325 days ago | |

I don't think smart ones are that fast. They can work for hours if you have the budget.

remram 324 days ago | | |

Interesting, I didn't know that.

gpm 325 days ago | |

Currently - claude code is pretty slow. I've definitely had it take >15 minutes on the (faster than opus!) sonnet model just thinking and writing code without feedback or running long lasting tools. I expect this to change given that companies like cerebras exist and seem to know how to generate tokens in much less time, but the current state of the art is what it is.

Always - if you're going to pipe the result of some slow process back to them (like building a giant C++ project that takes minutes/hours, or running a huge set of tests...)... it's going to be slow.

louiskw 325 days ago | |

For me, the average task takes a coding agent 2-5 minutes to complete. A slightly annoying amount of time as I'm prone to getting distracted while I wait.

This gives me something to do in that time.

My guess is time to complete a task will oscillate - going up as we give agents more complex tasks to work on, and going down with LLM performance improvement.

broast 322 days ago |

This tool just made me realize I've been vibe coding at years for work as a manager. Great job, this is a game changer so far for me!

jjangkke 324 days ago |

I'm not sure if Kanban is the right UI for what this is supposed to be for, just a gut feeling. Curious what other UI is more appropriate for this.

louiskw 324 days ago | |

Kanban seems like a good starting place, but I broadly agree that the interface for human<>agent collaboration will need to be different from the default interface we have today with legacy PM tools.

Things move across the board so quickly when AI is doing the work that ~50% of the columns seem pretty redundant.

skeptrune 325 days ago |

I remember the original Bloop search engine!

Some kind of UI or management system like this seems like it would be high level useful. Will have to give it a run.

louiskw 325 days ago | |

Nostalgia!

Let me know if any issues, we're turning feedback around pretty quick

csomar 325 days ago |

Why does this need GitHub auth? This asks for unlimited private access to ones repo. This is a hard NO from me.

louiskw 324 days ago | |

It can open GitHub PRs from the interface and there's tons more info we want to pull in like the result of CI checks

csomar 324 days ago | | |

Then you should have gone with a GitHub App instead of an Oauth App. The difference is that a GitHub App allows granularity of selection.

MH15 325 days ago |

Can you use this with the Claude Code Github Action?

bpshaver 325 days ago |

> Get 10x more out of...

So you're saying it goes up to 11x?

louiskw 325 days ago | |

Great ChatGPT question XD

skeeter2020 324 days ago | |

AI was already giving you 100x so we're up to 1000x

drbojingle 324 days ago |

I'm building one of these too :D

shibeprime 325 days ago |

Now we just let a CTO agent create the cards, review, and merge the PRs?

ggordonhall 325 days ago | |

We have this! Vibe Kanban includes its own MCP server that you can use to create tickets within the Kanban board.

Click on the MCP Servers tab, then hit "Add Vibe Kanban MCP". Then create and start a "planning" ticket like "Plan a migration from AWS to Azure and create detailed tickets for each step along the way". Sit back and watch the cards roll in!

Will do more to document this better soon :)

amirhirsch 325 days ago |

great. now i have to spend my morning using vibe kanban to make a tui for vibe kanban with textual. i'll submit a pr as soon as it's finished.

Jolter 325 days ago | |

It says in the linked readme to not open pull requests without discussing it first.

lowdownbutter 325 days ago |

[flagged]