Hi! Co-founder here. Conductor actually kicked off with a HN post many lifetimes ago in July 2025 (https://news.ycombinator.com/item?id=44594584). I owe so much to this community. Thank you all so much!
The whole team is around to help answer any questions. You can reach out to me personally at charlie [at] melty.sh.
I'm in the awkward position of wanting to congratulate you and scold you at the same time. Wish you'd been bought by Anthropic or something instead. Anyhow- congrats on getting scooped up.
To my understanding Muse has finally caught up to Anthropic, OAI and others atleast on coding and other benchmarks. This move makes the most sense to now quickly try to capture the market.
im so curious what specifically makes this possible. I have no experience with conductor but your comment reminds me of how impressed I was with the development speed of parse.com. I'm likely dating myself; it was the first serverless/firebase backend concept that supported mobile app development. I wasn't even a user I just noted how prolific the feature rollouts were. So polished too. They got acqui-hired by Meta and I'm sure the founders did well.
Anyway, for anyone reading, just a not developer-led products. How do they do it?!
Just wanted to congratulate you guys. Like many of us here, I had a mixed reaction of happy for your team, but also sad for Conductor. I tried a bunch of different approaches for coding agents, once I started using Conductor I was like "oh I guess I'm done looking for a harness" - a beautiful clean UI, great UX and just an overall smooth experience with all the features you could need.
Hopefully Meta does something good with your team's talent and technology, but regardless, it was a huge productivity boost for me and I really appreciate that. Good luck to you and your team at Meta!
I really fail to understand what is going on here.
There is literally no moat in most software expect the users a product has acquired. The number of slightly different solution to an existing product is wild. And this is getting crazier. I build for the fun of it now. Completely lost any sight of $ goals.
I am building agents, have own gui framework, own server orchestration, and now looking at deploying multi-app servers using unikernels. I build for my niche audience and make them happy - I am going via my existing clients who are in the property management domain. Non technical folks who have to handle everything from inspection to finance. My ideal type of folks.
I am sure by the time I build my end to end agent/orchestration/deployment/management system, I will have some customers and many more clones. Slightly different clones. This is the future. Why would anyone acquire anything?
Companies acquire talent
;Companies acquire products with a fast-growing usage base
;Companies acquire products that can help them beat competition in a fraction of the time it would take to build internally
(in decreasing order of "real value" and increasing order of self-deception)
I feel its less about the tech and more about the stage Meta and Conductor are in, right now
That's not necessarily true. They can be using it to jumpstart their own desktop app.
I don't see a Muse desktop app, this gives them something they can spin up in a few weeks, not months, with native cloud integration. I'd guess we're going to see a product competing with OpenAi/Anthropic (as far as standalone apps are concerned) by January.
You mean that desktop app is not something I can spin up in days with my tiniest of LLM budgets?
I do not know where people are lost these days. I build ambitious apps with 3x $20 subscriptions. And I live in a village in eastern Himalayas. People are really in a bubble.
LLMs have crushed the idea of "using [another product] to jumpstart" anything. Just build it.
Yes you can build it in a single prompt, but you miss the 93 versions of iterations and thoughts of the developers who've already made the project. If you have the money, why not buy a product/team that already has a year of experience with real users instead of reinventing the wheel?
That is your assumption. I am not talking about one-shot. I mean that lots are people are going through more iterative development cycles than ever in the history of software. Just my own examples (100% agent written)
All these and more, from a single engineer with 3 x $20 subscriptions. While I also run chicken and pig farms. It will take time for people outside tech hubs to understand what LLMs enable.
Not that I don’t think there’s amazing things that can be built easier, but designing and building a product is more than writing the code. Look at where you spent effort for your own examples. There’s a skill there which goes further than just reverse engineering an existing solution, and that skill is obviously of value to Meta. People were going crazy for the price they paid for instagram some years ago I remember.
I mean okay but you're looking at two very different things. At your description, it makes no sense to acquire your business. But Conductor is not building for a niche audience and is a radically different beast. Even before this acquisition they raised over 20 mil. You're using very naive thinking when you're talking about value here, its not as simple as why buy a company when someone can vibecode it. This is particularly funny because its usually AI labs that are following through with some of the biggest buys lately.
Why do you think how I am building is not going to be the norm? Software is going the way of plumbing. Pick a tiny niche, build highly customized workflows and sell to them. Global products finding audience across 100 countries is no longer working.
Raising money was not an indicator of value before LLMs but it helped a lot. I have been tech lead of a couple US based seed-funded startups and I definitely see the value.
But most software has no intrinsic hard engineering these days. A few do. Companies raise money because they feel they will grow into a large user-base. Then, either make money or get acquired. Where is "value" here? Its VC-speak all the way.
Privately funded AI labs in the US are buying, yes. I highly doubt this trend will go on when the return on investment starts fading. There is momentum in executives. The "value" is how they used to see things. Takes time to update.
Wednesday, Nov 4 New users can no longer sign up for Conductor or download the app.
Friday, Nov 13 Conductor Cloud goes read-only
Local Conductor will continue working as normal for now. It will start winding down in early 2027.
If you’re interested in self hosting your own, I’m the creator of woltspace [0], it’s a fully open-source way of managing your team of coding agents (wolts). You give the wolts a lodge on your own infra. They’re always on, always reachable, and fun beavers and racoons that evolve over time. Happy to chat about it! It’s by far my favourite workflow as a dev who loves tmux and being in a terminal
(disclosure: I'm a founder of yet another T3/Conductor competitor, https://boxes.dev)
You might want to do a quick eval of all the products in the space vs your org's needs! There are a lot of options now, and T3 isn't really a drop-in replacement for Conductor if you wanted the cloud setup.
Specifically T3's great for local, but for cloud they don't host anything - you just run their server on your own machine / VM. Which works ok, esp if you prefer full self-hosted, but then parallel agents still share a machine and need worktrees for isolation. Conductor Cloud actually eliminated worktrees by giving each workspace its own VM with a dev environment preloaded - which is the same approach we (and a few others) are taking.
Also you're not mentioning that Boxes is not open-source, where as T3 is. They just had to leave Conductor essentially because it was an acquihire of a closed source offering. I think if we're to stay safe in an environment that's moving this fast: we do need the code so that we can continue using it in case of these M&A's.
Also - if you are a developer/SWE this 'keep your laptop closed' thing with these control planes is mostly hype. I'd place my bets on something like Rex (https://www.superlogical.com/) instead.
We've been building https://www.superconductor.com/ since before Conductor had cloud (actually, technically before Conductor was released, so the naming is a funny coincidence). Our team's been working together for over a decade across multiple startups (and one prior exit) and we put pride into building an excellent multiplayer, multi-surface, mobile-friendly experience so teams can work together with agents anywhere on any project. Give us a look and let us know how it goes!
We built https://boxes.dev specifically for cloud dev / isolated VMs (disclosure, I'm one of the founders). Getting work off your laptop is such a productivity boost when you have parallel threads. Obviously there are a ton of products in this category (rest of the comments here have many examples) but we're 100% focused on the cloud experience
Trying this out... based on Claude's assessment it's got the fullest support for my existing workflow requirements compared to some of the others ones suggested.
I will always recommend T3 Code, it is by far the best one I have tried that isn't glued to a specific model provider. Trades blows with Codex / Claude Desktop in terms of features. All of the other multi-provider orchestrators have not been keeping up with Claude Desktop in particular adding features at lightning-speed, but T3 Code somehow manages.
(I am building my own tool in this area and I literally downloaded and tried all of them as market research)
I switched from conductor to orca because conductor didn’t support mobile app working with local instances and didn’t have a good cli/mcp at the time.
Very happy I switched. I’m extremely productive on mobile. It defiiitely feels power usery but i am one so it works great.
My only gripe is sometimes it loses connections to workspaces in a way that is hard to recover BUT the meta conductor living on top can resolve it for me. Kudos for having it.
appImage is not a great packaging format for Linux. Any chance you could ship a flatpak instead? You'd likely need to host your own flatpak remote, but it's not rocket surgery.
Thank you for your consideration and for your work.
I'm one of the founders of https://podium.do. We created Podium because none of the other ADEs worked like what we wanted. I think it could be interesting for you to checkout as an alternative because there are similarities to Conductor, plus a few interesting additional features. We are pretty strong for complex tasks because we put an agent native issue tracker at the core of architecture which turned out as a super power in that case.
Free 3 month sub to any current Conductor user. We promise to never sell to Facebook, X or other weirdo billionaires. I started this company to get away from working at big tech and not to sell to it. Up to 10 cloud workspaces per person, works with all the things you need.
https://hive.codes/signup
Give my dreamcoder.ai harness a whirl. Runs in your browser on all 3 platforms. Tiling window manager; generous use of keyboard shortcuts, so it's not just for mouse users.
Is anybody else running into the problem trying to sign up for Conductor, where it says that new signups are closed and offers you no other route forward? There are some docs referencing the Meta acquisition and a focus on local since they're closing cloud, but no way to actually use the Mac app locally without signup.
It seems like it's got a lot of utility for my team and I, but jesus
Their announcement literally says that Conductor Local will be supported into 2027, which is 2 months away. Now is not the time to be jumping on board with it.
Well, if any of the disgruntled people on this thread fancy moving to an agent with a similarly-themed name and vibe, Juggler (https://github.com/juggler-ai/juggler) is available.
No sign-up required - the angle I'm taking is that juggler is a client-server architecture, and if you want cloud VM sessions, you just spin up your own server and run a juggler server process on there. Its GUI is identical whether it's a local or remote session.
(I'd never actually heard of conductor, but interesting to see that people do want that kind of managed cloud backend server. I might think about the logistics of offering that)
And I promise to never sell-out to Meta[1].
[1] Nah, of course I would if they offered me a really huge amount of money.
Congratulations to the team - but it's a bitter ending. I wish, in these cases, the company could open source a version of their desktop software. But I'm sure the tech is part of what Meta is acquiring.
I started using Conductor in January this year and it was a great tool - Charlie and the team knocked it out of the park. I only migrated to Orca when I needed better multi-machine control, and remote control, but these tools are a dime-a-dozen now. I'm not sure that if I was hunting for a new tool today that I'd pick up Orca. Maybe Herdr? or Stencil's Tern?
All that to say - they built a great product with great UX. Looking forward to seeing what they build next at Meta!
My needs (multi machine set up, one tool) do not work well with “one VM”. If I ran Claude code on one VM, I’d only have that one machines resources. I’d have to have a second session in another terminal tab which is… well that’s Herdr. And orca. They manage those.
I was just commenting yesterday (https://news.ycombinator.com/item?id=50367546) how I really like Conductor except that it doesn't work so well if you want to work across multiple repositories.
I guess that's the kick I needed to dig into the alternatives.
I switched to Paseo long ago because of its full featured mobile app, local client server architecture, and open sourcedness. Conductor began to feel stale, pricey, and understandably but annoying for me moving toward enterprise.
This morning my team was a bit grumpy about this. This afternoon we've already forked + reverse engineered the whole project, including Passport (our favorite feature)! LLMs truly are amazing, lol.
Shameless plug: I made https://aweb.ai, which does essentially the same but can also add chatgpt and claude.ai instances to the team, works for agents across many computers and organizations, lets you organize them in teams, gives your agents private or public ids (multi-register and federated), and is OSS.
Why not try one of the existing services before you spend time maintaining something yourself? For the past two months, I've been using boxes.dev
95% of all my sessions are on boxes. Initially, the draw was just a remote box that I could control on mobile, but some of the features they have added recently, like coordinating child sessions, are wildly good.
I know there are new startups in this space every day, so what tipped things for me was that the team is so responsive and ships features in the matter of hours.
I'll tinker with this on the side. To see "how hard can it be anyhow" and to maybe end up with something better long term.
-----
Edit:
Surprise, surprise - I'm launching a new open source project: "Conducted" (not affiliated with, or to be confused with any similar brands)
Conducted is MIT license, and will remain open source until hell has frozen over.
Looking for contributors ;) .. Feel free to give it a star, throw some agents in its direction.
I was a early user of conductor, even gave a lot of feedback through the inapp button (the team responded to me personally and quickly always which was nice, and my feature requests got into the app quickly). I've stopped using it a long time ago but at the time it was a game changer for me.
The economics of these harnesses I don't think lend to many winners, the model usage plans the big guys have at offer will always undercut the small players no matter how much they try to resell an openmodel for $20/mo. Congrats to the team though!
Site looks great, congrats - will have a peek - but I feel task handling is better done in the repo itself? (repo-as-the-harness thinking) especially for teams.
Interesting idea, but we find task data is too high throughput to put into git. I assume with the new crop of DVCSs that can handle much more volume this might change.
I had the same question. I think this is in the space of being like the 'trillian' or 'matrix' of ai agents, using multi provider coding agent from a single interface plus some bells and whistles.
But I think within claude code and codex subagent support has improved a lot lately and it doesn't seem this would improve that much over that native support if you only use one provider.
The benefit is not spinning up sub-agents, but spinning up parallel workstreams. They basically took 90% of the friction out of working with and using worktrees. I was using Linear a lot (and enjoying it!) for my issue tracking, but at some point I realized it's more productive to just create a new worktree in Conductor for any bug/feature and cut out the middleman.
hmm so working on different features let's say on the same git, in parallel, without causing issues? Is the improvement of this parallelization worth it on solo projects or it becomes only relevant on big projects with lots of PRs and such?
It's always useful to have even if you don't need 8 worktrees with agents all going at the same time (I rarely have more than 2-3 active at a time). Typically there a tension between the 'fast way' (a big ol' slop branch bag of assorted changes) and the 'correct way' (carefully isolated branches and PRs) - with a smart implementation of worktrees like Conductor, the correct way becomes just as fast. That's basically the heart of good UX in my mind.
If you’re looking for an open-source alternative, I’m the creator of woltspace [0], which is a way of coordinating your wolts (coding agents) by letting them talk to each other and giving them a lodge on your machine.
I built it for myself as a dev who loves tmux and vim and who was very unsatisfied with the typical agent orchestration experience. You should check it out! The biggest advantage is you bring your coding agents to woltspace, not the other way around.
Western AI giants have hundreds of billions of dollars to make back in a world where China is releasing open source tooling and open weights models. How else does they make their money back other than finding _something_ proprietary to sell.
I switched to Paseo a few months ago. Fully open-source, I've had no issues with it. But no cloud provider option, so to speak. (I run an instance of Paseo in a Hetzner box, agents working full-time, haven't really had to do any maintenance on it...)
FWIW I used Orca for a week and then went back to Herdr..
Orca has a lot of UI bugs, buttons that don't work, menus that don't disappear, etc.. To be fair they push updates quite regularly, but I think I will check back in once it's a bit more stable.
If you like taking bets on smaller oss & independent projects that have no interest in creating a vc backed business, I'm working (9 months+ now) on https://ouijit.com/
For what it's worth, I'm using plain Claude Code: one git worktree per session, the sessions messaging each other directly, and one coordinator session that assigns the work. To be honest, it fills the multi-agent gap pretty well. It's not as polished as Conductor, but given how much Claude Code itself has improved over the last few months, and where it's heading, I think it's a good alternative to just stick with.
I used conductor when it was first released for quite some time. Great product but i ultimately wanted to use my own networked devices to do work, so i moved to herdr and then ultimately to t3 code now.
Very happy to see the team get a happy ending/landing!
Congrats to the team! This is a great acquire for meta to super speed their coding agent rival. It's very exciting to now have another tier 1 major player in the space.
I created https://superintelligencehive.com/ because I wanted my own orchestrator - it is local and has most of the features you could want. And forever free.
Your website focuses quite a bit on speed. Can you talk about _how_ it achieves this (besides potentially running on a server that's closer to where the models are hosted?)
just when I was starting to really like conductor. It's a really good app / idea for working with agents. The strongest workflow I've used. Ig I'll move to herd now.
nice to get paid by meta, but cmmon read the room.
woahhhh congrats to the conductor team! If folks are looking for cloud agents/replacement, can't recommend boxes.dev enough, they've got phenomenal support and ship INCREDIBLY fast.
Tried Conductor in the past but had migrated all of our teams remote development work to http://boxes.dev as it was much more intuitive and easier to manage parallel workstreams.
Would highly recommend it for anyone looking for an alternative with Conductor shutting down.
every time a new tool appears, everyone around me is like "you should try you should try it". but i am here, stuck with my vim and tmux. and im glad i did. most of the new shiny things get enshittifed or acquired by shady corporations. long live vim+tmux
well, conductor was so great that after using it for a few weeks, I completely stopped fearing a lock-in. I just use the official cli + desktop apps now. tmux + a bit of custom scripting works wonders.
(As someone who isn't particularly happy about this news)
> why sellout your company and your customers and community?
Why not? Do you think this is a heart project, that the founders have a personal passion for agent orchestration somehow? They're not curing cancer or supporting orphans, most companies are created to make money, that's how our economic system works (for better or worse, but I'm not going to blame individuals or be surprised about it).
"community"... Come on. There's customers, there's no community here.
Realistically for them, coding harnesses are an extremely competitive landscape, and they have to live in constant fear that Anthropic/OpenAI etc will rugpull them and force these harnesses to go through API billing (effectively killing it as an option for I would guess the majority of their users). That nearly happened earlier this year when Anthropic intimated they were going to do this.
The other reality is, there are now good free alternatives to Conductor (I haven't tried them so we will see) - it does suck if you feel very tied to your existing workflow, but I also think about it from the level of 'these people gave me something for free that helped me a lot, and I want things to work out well for them'.
Hi! Co-founder here. Conductor actually kicked off with a HN post many lifetimes ago in July 2025 (https://news.ycombinator.com/item?id=44594584). I owe so much to this community. Thank you all so much!
The whole team is around to help answer any questions. You can reach out to me personally at charlie [at] melty.sh.
"... many lifetimes ago in 2025"
This made me laugh. The world is different now. Congrats!
I'm in the awkward position of wanting to congratulate you and scold you at the same time. Wish you'd been bought by Anthropic or something instead. Anyhow- congrats on getting scooped up.
What value could they possibly provide Anthropic? This is by far the best-case scenario for the team
A great, stable product - that doesn't try to be absolutely everything - but does a few things really well, and looks great while doing it?
That's not Anthropic.
Conductor is really no different from Claude Code Desktop now, aside from multiple harness and models
To my understanding Muse has finally caught up to Anthropic, OAI and others atleast on coding and other benchmarks. This move makes the most sense to now quickly try to capture the market.
You lads... I remember when we had to paste the code from the chat window into VSCode.
Remember when 200k was a HUGE context window, but you still had to be precious about how many files you'd paste into it?
Are you going to retire now and start a band ?
Congratulations, getting an exit in a bit over a year is very very good
Best shipping speed I've experienced for a product like this. Great run!
Consider Open Sourcing the core/app
im so curious what specifically makes this possible. I have no experience with conductor but your comment reminds me of how impressed I was with the development speed of parse.com. I'm likely dating myself; it was the first serverless/firebase backend concept that supported mobile app development. I wasn't even a user I just noted how prolific the feature rollouts were. So polished too. They got acqui-hired by Meta and I'm sure the founders did well.
Anyway, for anyone reading, just a not developer-led products. How do they do it?!
Not a user but congrats and I hope you make fat stacks.
I loved the autographs at the bottom. It gives a nice feeling. Way to go!
Congrats! This is only the beginning of your journey!
Just wanted to congratulate you guys. Like many of us here, I had a mixed reaction of happy for your team, but also sad for Conductor. I tried a bunch of different approaches for coding agents, once I started using Conductor I was like "oh I guess I'm done looking for a harness" - a beautiful clean UI, great UX and just an overall smooth experience with all the features you could need.
Hopefully Meta does something good with your team's talent and technology, but regardless, it was a huge productivity boost for me and I really appreciate that. Good luck to you and your team at Meta!
congrats to the team!
Congrats Charlie!
I really fail to understand what is going on here.
There is literally no moat in most software expect the users a product has acquired. The number of slightly different solution to an existing product is wild. And this is getting crazier. I build for the fun of it now. Completely lost any sight of $ goals.
I am building agents, have own gui framework, own server orchestration, and now looking at deploying multi-app servers using unikernels. I build for my niche audience and make them happy - I am going via my existing clients who are in the property management domain. Non technical folks who have to handle everything from inspection to finance. My ideal type of folks.
I am sure by the time I build my end to end agent/orchestration/deployment/management system, I will have some customers and many more clones. Slightly different clones. This is the future. Why would anyone acquire anything?
They need my prompts, not my tech.
Companies acquire talent ;Companies acquire products with a fast-growing usage base ;Companies acquire products that can help them beat competition in a fraction of the time it would take to build internally (in decreasing order of "real value" and increasing order of self-deception)
I feel its less about the tech and more about the stage Meta and Conductor are in, right now
In this case they're shutting down the product, so it's definitely not for the product or for the user base.
That's not necessarily true. They can be using it to jumpstart their own desktop app.
I don't see a Muse desktop app, this gives them something they can spin up in a few weeks, not months, with native cloud integration. I'd guess we're going to see a product competing with OpenAi/Anthropic (as far as standalone apps are concerned) by January.
100%. I'm seeing it as a repeat of the VS Code fork wars. Funnily enough, nobody won that war. The paradigm disappeared entirely.
You mean that desktop app is not something I can spin up in days with my tiniest of LLM budgets?
I do not know where people are lost these days. I build ambitious apps with 3x $20 subscriptions. And I live in a village in eastern Himalayas. People are really in a bubble.
LLMs have crushed the idea of "using [another product] to jumpstart" anything. Just build it.
Yes you can build it in a single prompt, but you miss the 93 versions of iterations and thoughts of the developers who've already made the project. If you have the money, why not buy a product/team that already has a year of experience with real users instead of reinventing the wheel?
That is your assumption. I am not talking about one-shot. I mean that lots are people are going through more iterative development cycles than ever in the history of software. Just my own examples (100% agent written)
https://github.com/brainless/nocodo, https://github.com/brainless/dwata, https://github.com/brainless/akar, https://github.com/brainless/baho
All these and more, from a single engineer with 3 x $20 subscriptions. While I also run chicken and pig farms. It will take time for people outside tech hubs to understand what LLMs enable.
Not that I don’t think there’s amazing things that can be built easier, but designing and building a product is more than writing the code. Look at where you spent effort for your own examples. There’s a skill there which goes further than just reverse engineering an existing solution, and that skill is obviously of value to Meta. People were going crazy for the price they paid for instagram some years ago I remember.
These were true in pre-LLM era. Cursor is a good example. Big numbers but what is the value in a year?
Sure they had a lot of growth but when there are a 100 alternatives, each offering something a bit different, then user bleed is natural.
Meta, Google and all big co will spend more $ short-term to buy and they will continue to bleed till the new playbook sets in the minds of execs.
talent it is then;
both acquihires and products will get cheaper, could the new playbook be the OG marketplace ?
I mean okay but you're looking at two very different things. At your description, it makes no sense to acquire your business. But Conductor is not building for a niche audience and is a radically different beast. Even before this acquisition they raised over 20 mil. You're using very naive thinking when you're talking about value here, its not as simple as why buy a company when someone can vibecode it. This is particularly funny because its usually AI labs that are following through with some of the biggest buys lately.
Why do you think how I am building is not going to be the norm? Software is going the way of plumbing. Pick a tiny niche, build highly customized workflows and sell to them. Global products finding audience across 100 countries is no longer working.
Raising money was not an indicator of value before LLMs but it helped a lot. I have been tech lead of a couple US based seed-funded startups and I definitely see the value.
But most software has no intrinsic hard engineering these days. A few do. Companies raise money because they feel they will grow into a large user-base. Then, either make money or get acquired. Where is "value" here? Its VC-speak all the way.
Privately funded AI labs in the US are buying, yes. I highly doubt this trend will go on when the return on investment starts fading. There is momentum in executives. The "value" is how they used to see things. Takes time to update.
Nobody's wrong or right. It's just comparing apples to bananas.
Things will shut down over the next few months. https://www.conductor.build/announcement:
Flashback to Cloudflare acquiring Deno and hiding the shutdown at the bottom of their announcement -- https://news.ycombinator.com/item?id=50019911.I think this is the sad part about it.
I wonder what would be the way to protect OSS like this from this kind of end?
It's a great win for the team, but many will be looking into other solutions for a while.
Months? More like weeks.
Literally just went to go "all in" on conductor yesterday, and buy into their cloud setup and hope the company keeps going.
And today we're doing a refund + moving the org to t3.codes instead.
Boo Meta. Boooooooo.
It is kind of interesting on how only the buyer gets booed.
If you’re interested in self hosting your own, I’m the creator of woltspace [0], it’s a fully open-source way of managing your team of coding agents (wolts). You give the wolts a lodge on your own infra. They’re always on, always reachable, and fun beavers and racoons that evolve over time. Happy to chat about it! It’s by far my favourite workflow as a dev who loves tmux and being in a terminal
[0] https://github.com/jerpint/woltspace
(disclosure: I'm a founder of yet another T3/Conductor competitor, https://boxes.dev)
You might want to do a quick eval of all the products in the space vs your org's needs! There are a lot of options now, and T3 isn't really a drop-in replacement for Conductor if you wanted the cloud setup.
Specifically T3's great for local, but for cloud they don't host anything - you just run their server on your own machine / VM. Which works ok, esp if you prefer full self-hosted, but then parallel agents still share a machine and need worktrees for isolation. Conductor Cloud actually eliminated worktrees by giving each workspace its own VM with a dev environment preloaded - which is the same approach we (and a few others) are taking.
Also you're not mentioning that Boxes is not open-source, where as T3 is. They just had to leave Conductor essentially because it was an acquihire of a closed source offering. I think if we're to stay safe in an environment that's moving this fast: we do need the code so that we can continue using it in case of these M&A's.
Also - if you are a developer/SWE this 'keep your laptop closed' thing with these control planes is mostly hype. I'd place my bets on something like Rex (https://www.superlogical.com/) instead.
[flagged]
Theo's awesome!
Theo is the reason I signed in to youtube on my living room TV so I can block his channel.
He is just annoying, otherwise he is fine.
As annoying as he can be, t3 code really speaks for itself, especially with the later updates
He's a narcissist
Extremely bummed about this. Having cloud webhooks + workspace/session API + isolated VMs has been a massive productivity boost for me.
Does anyone have suggestions for products that can match Conductor Cloud, not just Conductor Local?
We've been building https://www.superconductor.com/ since before Conductor had cloud (actually, technically before Conductor was released, so the naming is a funny coincidence). Our team's been working together for over a decade across multiple startups (and one prior exit) and we put pride into building an excellent multiplayer, multi-surface, mobile-friendly experience so teams can work together with agents anywhere on any project. Give us a look and let us know how it goes!
We built https://boxes.dev specifically for cloud dev / isolated VMs (disclosure, I'm one of the founders). Getting work off your laptop is such a productivity boost when you have parallel threads. Obviously there are a ton of products in this category (rest of the comments here have many examples) but we're 100% focused on the cloud experience
Trying this out... based on Claude's assessment it's got the fullest support for my existing workflow requirements compared to some of the others ones suggested.
Awesome, thanks a ton for trying it! Definitely reach out of you hit any issues or have feedback, email is drew at boxes.dev
Orca dev here! I'd suggest Orca
Why? We let you bring your own cloud (versus being locked in).
You can bring a local cloud, sandboxes, or whatever you'd like
We even launched community plugins this week for VM/sandboxes (our you can have your own internal plugin)
https://www.onorca.dev/plugins?kind=machines
Paradev founder here, our cloud workspaces should cover most of that https://paradev.com
give https://paseo.sh a go, self host it to get many of the same features and more
+1 - we've been very happy on Paseo.
We like the web + mobile app orientation (available without an app store!) along with the high degree of control.
Anyone moved between Paseo and T3? Curious to hear what motivated you in either direction.
I'd say tern or orca, some customization and a decent vps will get you 95%+ of the way there
superset.sh is quite a good option that will mostly get you there.
I switched to T3 codes a few weeks ago after being a power user of Conductor and I understand why the product felt a bit stale now.
What do you get out of T3? What's the appeal?
Heavy conductor user here. Does the hacker news community have any suggestions for replacements?
I will always recommend T3 Code, it is by far the best one I have tried that isn't glued to a specific model provider. Trades blows with Codex / Claude Desktop in terms of features. All of the other multi-provider orchestrators have not been keeping up with Claude Desktop in particular adding features at lightning-speed, but T3 Code somehow manages.
(I am building my own tool in this area and I literally downloaded and tried all of them as market research)
Wait, there's something called Claude Desktop that's actually useful for non-non-technical people?
I second this. T3 code especially for remote machine workflows is unmatched.
https://orca.dev I think it's probably the easiest switch (Conductor imports).
I work on Orca and we're shipping frequently.
Let us know if you have feedback! (MIT OSS)
I switched from conductor to orca because conductor didn’t support mobile app working with local instances and didn’t have a good cli/mcp at the time.
Very happy I switched. I’m extremely productive on mobile. It defiiitely feels power usery but i am one so it works great.
My only gripe is sometimes it loses connections to workspaces in a way that is hard to recover BUT the meta conductor living on top can resolve it for me. Kudos for having it.
If you make a GH issues for your connection problem we can also probably get to the bottom of it!
appImage is not a great packaging format for Linux. Any chance you could ship a flatpak instead? You'd likely need to host your own flatpak remote, but it's not rocket surgery.
Thank you for your consideration and for your work.
It is far, far easier to convert an AppImage into a Flatpak (or any other packaging form) than vice versa.
If a dev is only going to choose one, I'd much rather they ship an AppImage.
there are so many and I'm sure something has changed since I last checked but:
- paseo: open, mobile app is a first class citizen, i've been happy with it
- orca: vc backed, featureful (built in ios sim, for example)
- herdr: vc backed, terminal native, im sure there are more features since i last checked it
- t3: open, never used it but folks seem to like it
I moved to Paseo. Open source and let's you drive from your phone.
https://herdr.dev/
I'm one of the founders of https://podium.do. We created Podium because none of the other ADEs worked like what we wanted. I think it could be interesting for you to checkout as an alternative because there are similarities to Conductor, plus a few interesting additional features. We are pretty strong for complex tasks because we put an agent native issue tracker at the core of architecture which turned out as a super power in that case.
Does Podium support Gemini?
Free 3 month sub to any current Conductor user. We promise to never sell to Facebook, X or other weirdo billionaires. I started this company to get away from working at big tech and not to sell to it. Up to 10 cloud workspaces per person, works with all the things you need. https://hive.codes/signup
Founder demo https://www.youtube.com/watch?v=pghD8n_Prw4 :-)
I use Paradev (https://paradev.com - previously Superset) and it's great. They posted a migration guide today https://docs.paradev.com/migrate-from-conductor
https://ampcode.com/ is amazing! Surprised the community here does not discuss it more.
Could be a Laravel and PHP X bubble thing, former JS/TS only devs mostly seem to promote T3.
https://imbue.com/product/sculptor
https://utensils.io/claudette/
Give my dreamcoder.ai harness a whirl. Runs in your browser on all 3 platforms. Tiling window manager; generous use of keyboard shortcuts, so it's not just for mouse users.
I like using https://paradev.com/
highly recommend https://www.scape.work - it's unlimited agents in parallel, and unlimited orchestrators in parallel.
Check Beolis (https://beolis.com).
Disclaimer: I'm developing it.
Agentastic is great for mac users. I'm hoping the project gets a boost with this acquistion.
I'm obviously biased but Paradev, it imports your Conductor worktrees as is https://paradev.com
Paseo is great
tern or orca or bb or vanilla claude code desktop / codex desktop
Claude or ChatGPT
boxes.dev!
getbb.app
macro.land
buzzpot.app :)
Is anybody else running into the problem trying to sign up for Conductor, where it says that new signups are closed and offers you no other route forward? There are some docs referencing the Meta acquisition and a focus on local since they're closing cloud, but no way to actually use the Mac app locally without signup.
It seems like it's got a lot of utility for my team and I, but jesus
Their announcement literally says that Conductor Local will be supported into 2027, which is 2 months away. Now is not the time to be jumping on board with it.
Well, if any of the disgruntled people on this thread fancy moving to an agent with a similarly-themed name and vibe, Juggler (https://github.com/juggler-ai/juggler) is available.
No sign-up required - the angle I'm taking is that juggler is a client-server architecture, and if you want cloud VM sessions, you just spin up your own server and run a juggler server process on there. Its GUI is identical whether it's a local or remote session.
(I'd never actually heard of conductor, but interesting to see that people do want that kind of managed cloud backend server. I might think about the logistics of offering that)
And I promise to never sell-out to Meta[1].
[1] Nah, of course I would if they offered me a really huge amount of money.
Judging by the number of alternatives that are promoted by their founders here there is clearly no moat in this
Congratulations to the team - but it's a bitter ending. I wish, in these cases, the company could open source a version of their desktop software. But I'm sure the tech is part of what Meta is acquiring.
I started using Conductor in January this year and it was a great tool - Charlie and the team knocked it out of the park. I only migrated to Orca when I needed better multi-machine control, and remote control, but these tools are a dime-a-dozen now. I'm not sure that if I was hunting for a new tool today that I'd pick up Orca. Maybe Herdr? or Stencil's Tern?
All that to say - they built a great product with great UX. Looking forward to seeing what they build next at Meta!
why use a 3rd party tool instead of just spinning up a VM in GCP and calling it a day? Would really appreciate your insight
My needs (multi machine set up, one tool) do not work well with “one VM”. If I ran Claude code on one VM, I’d only have that one machines resources. I’d have to have a second session in another terminal tab which is… well that’s Herdr. And orca. They manage those.
I was just commenting yesterday (https://news.ycombinator.com/item?id=50367546) how I really like Conductor except that it doesn't work so well if you want to work across multiple repositories.
I guess that's the kick I needed to dig into the alternatives.
Great exit in a crowded space.
I switched to Paseo long ago because of its full featured mobile app, local client server architecture, and open sourcedness. Conductor began to feel stale, pricey, and understandably but annoying for me moving toward enterprise.
This morning my team was a bit grumpy about this. This afternoon we've already forked + reverse engineered the whole project, including Passport (our favorite feature)! LLMs truly are amazing, lol.
care to open source it ?
Shameless plug: I made https://aweb.ai, which does essentially the same but can also add chatgpt and claude.ai instances to the team, works for agents across many computers and organizations, lets you organize them in teams, gives your agents private or public ids (multi-register and federated), and is OSS.
Conductor is awesome. Meta sucks donkey balls, and I refuse to use their products.
I'll prolly have claude reverse engineer conductor over the next month so the show can go on ...
There’s no need to. There are so many alternatives with almost the same feature set like superset.sh and orca.dev.
Why not try one of the existing services before you spend time maintaining something yourself? For the past two months, I've been using boxes.dev
95% of all my sessions are on boxes. Initially, the draw was just a remote box that I could control on mobile, but some of the features they have added recently, like coordinating child sessions, are wildly good.
I know there are new startups in this space every day, so what tipped things for me was that the team is so responsive and ships features in the matter of hours.
I'll tinker with this on the side. To see "how hard can it be anyhow" and to maybe end up with something better long term.
-----
Edit:
Surprise, surprise - I'm launching a new open source project: "Conducted" (not affiliated with, or to be confused with any similar brands) Conducted is MIT license, and will remain open source until hell has frozen over.
Looking for contributors ;) .. Feel free to give it a star, throw some agents in its direction.
https://github.com/jahala/conducted
do you work at boxes.dev
Nope, completely different space and zero affiliation. Which either means it's really that good, or that I need to start charging consulting fees.
Save tokens and spend them on building cool stuff instead. If you're open to it, give my dreamcoder.ai harness a whirl.
Agentastic is already pretty close in concept. Why not help improve it?
Been using superset for some time, but they dont have the cloud offering yet..
We have cloud now! also renamed to Paradev https://paradev.com
I migrated off Conductor to getbb.app a while ago and can highly recommend it.
For cloud agents I just built plugins in bb for boat.dev and smolmachines.com. Works like a charm.
literally no need when there are about 10,000 clones all doing the exact same thing, practically down to the UI and website
For anyone else who did a double take, this is different to Orkes Conductor which was the subject of a recent Jepsen analysis.
I was a early user of conductor, even gave a lot of feedback through the inapp button (the team responded to me personally and quickly always which was nice, and my feature requests got into the app quickly). I've stopped using it a long time ago but at the time it was a game changer for me.
Its crazy how fast the tech field is moving
The economics of these harnesses I don't think lend to many winners, the model usage plans the big guys have at offer will always undercut the small players no matter how much they try to resell an openmodel for $20/mo. Congrats to the team though!
I am pretty sure Conductor supported BYOAI since day 1. The monetization play was to charge on cloud boxes, not inference.
The changelog is truly hilarious:
> Conductor Cloud goes read-only on November 13 and shuts down on November 20.
3 sentences later
> Add your own completion sounds in Settings → General. Pick a sound file, name it, and choose it like any other sound.
If you're shutting down in *9 days* I really don't care that you added custom sound effects.
The number of comments in here saying “founder of xyz here, if you’re looking for a replacement we make a similar product” is illustrative.
Most people who need this product probably should just vibe code their own.
At least that's what I did.
Glad for the team, sad for the product
I'm not a Conductor user, but the shutdown timeline seems rather rushed no?
That's the consensus on the Discord server, people are not happy.
ok, as a fan of public transfer I just asked myself, how a conductor could be useful to Meta
Congrats to the team! It's great to see acquisitions in this space.
I personally never thought Conductor was that great, so I built my own with https://podium.do.
What I was missing was a built-in task system so more complex tasks can be split up by the agents. And running agents on my own machines of course.
Site looks great, congrats - will have a peek - but I feel task handling is better done in the repo itself? (repo-as-the-harness thinking) especially for teams.
Thank you!
Interesting idea, but we find task data is too high throughput to put into git. I assume with the new crop of DVCSs that can handle much more volume this might change.
Might be a dumb question but can you explain in a couple words the benefit VS a well organized claude code CLI that knows to spin sub agents?
I had the same question. I think this is in the space of being like the 'trillian' or 'matrix' of ai agents, using multi provider coding agent from a single interface plus some bells and whistles.
But I think within claude code and codex subagent support has improved a lot lately and it doesn't seem this would improve that much over that native support if you only use one provider.
The benefit is not spinning up sub-agents, but spinning up parallel workstreams. They basically took 90% of the friction out of working with and using worktrees. I was using Linear a lot (and enjoying it!) for my issue tracking, but at some point I realized it's more productive to just create a new worktree in Conductor for any bug/feature and cut out the middleman.
hmm so working on different features let's say on the same git, in parallel, without causing issues? Is the improvement of this parallelization worth it on solo projects or it becomes only relevant on big projects with lots of PRs and such?
It's always useful to have even if you don't need 8 worktrees with agents all going at the same time (I rarely have more than 2-3 active at a time). Typically there a tension between the 'fast way' (a big ol' slop branch bag of assorted changes) and the 'correct way' (carefully isolated branches and PRs) - with a smart implementation of worktrees like Conductor, the correct way becomes just as fast. That's basically the heart of good UX in my mind.
Gah this sucks! Conductor is my main development tool. Now I'm forced to look for alternatives. Why can't they open source it?
Over the years I've gotten extremely conservative with adopting any new tooling into my workflow, for exactly reasons like this.
If you’re looking for an open-source alternative, I’m the creator of woltspace [0], which is a way of coordinating your wolts (coding agents) by letting them talk to each other and giving them a lodge on your machine.
I built it for myself as a dev who loves tmux and vim and who was very unsatisfied with the typical agent orchestration experience. You should check it out! The biggest advantage is you bring your coding agents to woltspace, not the other way around.
[0] https://github.com/jerpint/woltspace
With labs spending their fundings left and right and vibe-coding, it's absolutely irresponsible to adopt anything less than 18 months old.
Trusts suck
They ruin innovation top down and bottom up at the same time
Not tried it yet but DeltaDB from ZED (and ZED editor more generally) could cover it for you and they are open source.
I switched to paseo.sh several months ago. It's just as good if not better.
respectfully, its a claude wrapper. with that said if you really can't live without it-- you have the binary, you know what to do.
Western AI giants have hundreds of billions of dollars to make back in a world where China is releasing open source tooling and open weights models. How else does they make their money back other than finding _something_ proprietary to sell.
Congrats to the conductor team! The timeline to migrate off Conductor Cloud feels very aggressive. What tooling are other folks moving to?
I switched to Paseo a few months ago. Fully open-source, I've had no issues with it. But no cloud provider option, so to speak. (I run an instance of Paseo in a Hetzner box, agents working full-time, haven't really had to do any maintenance on it...)
https://paseo.sh/
Orca is next for me to trial https://www.onorca.dev/
Herdr has been another YC favorite https://herdr.dev/
FWIW I used Orca for a week and then went back to Herdr..
Orca has a lot of UI bugs, buttons that don't work, menus that don't disappear, etc.. To be fair they push updates quite regularly, but I think I will check back in once it's a bit more stable.
Hey, if you have issues please let us know!
We test our product like crazy, so I'm quite surprised to hear this. But feel free to drop me a message, open a GH issue or otherwise.
We are OSS MIT and always open to feedback
If you like taking bets on smaller oss & independent projects that have no interest in creating a vc backed business, I'm working (9 months+ now) on https://ouijit.com/
gh: https://github.com/ouijit/ouijit
It's a bit unique from a design standpoint, but the next release brings it a little more in line with the zeitgeist (but not too much).
For what it's worth, I'm using plain Claude Code: one git worktree per session, the sessions messaging each other directly, and one coordinator session that assigns the work. To be honest, it fills the multi-agent gap pretty well. It's not as polished as Conductor, but given how much Claude Code itself has improved over the last few months, and where it's heading, I think it's a good alternative to just stick with.
We're building Datadog Studio[0], for both building things, but also following them through to prod
[0] https://www.projectcodename.dev/access
I moved to getbb.app. Been really happy with it for the past few months. The plugin functionality is really cool.
T3 Code here, I've been a hardcore user of Conductor for a while but T3 is excellent.
I used conductor when it was first released for quite some time. Great product but i ultimately wanted to use my own networked devices to do work, so i moved to herdr and then ultimately to t3 code now.
Very happy to see the team get a happy ending/landing!
Are you using Tailscale for your network, or something else?
I've been working on a startup in the "private networks + VMs for agents" space, so I'm always trying to learn how people run their current setups.
I first thought it was about this Conductor: https://github.com/netflix/conductor
Congrats to the team! This is a great acquire for meta to super speed their coding agent rival. It's very exciting to now have another tier 1 major player in the space.
I created https://superintelligencehive.com/ because I wanted my own orchestrator - it is local and has most of the features you could want. And forever free.
Congrats to the team, but bummed out that I'm losing my favorite IDE!
For anyone looking for a slack-native conductor check out opencomputer.dev
disclosure: I am one of the founders
Your website focuses quite a bit on speed. Can you talk about _how_ it achieves this (besides potentially running on a server that's closer to where the models are hosted?)
I will miss it dearly.
just when I was starting to really like conductor. It's a really good app / idea for working with agents. The strongest workflow I've used. Ig I'll move to herd now.
nice to get paid by meta, but cmmon read the room.
woahhhh congrats to the conductor team! If folks are looking for cloud agents/replacement, can't recommend boxes.dev enough, they've got phenomenal support and ship INCREDIBLY fast.
Tried Conductor in the past but had migrated all of our teams remote development work to http://boxes.dev as it was much more intuitive and easier to manage parallel workstreams.
Would highly recommend it for anyone looking for an alternative with Conductor shutting down.
every time a new tool appears, everyone around me is like "you should try you should try it". but i am here, stuck with my vim and tmux. and im glad i did. most of the new shiny things get enshittifed or acquired by shady corporations. long live vim+tmux
Everything you add to your tech stack is a liability. Period. I've always leaned minimalist.
well, conductor was so great that after using it for a few weeks, I completely stopped fearing a lock-in. I just use the official cli + desktop apps now. tmux + a bit of custom scripting works wonders.
Dang… I didn’t even get a chance to try it.
Glad I started migrating all my projects to Paseo yesterday!
"We launched Conductor a lifetime ago, in the distant past of July 2025. " Dear god the sheer smugness. This bubble can't burst soon enough.
The end.
> If you liked Conductor, you’re gonna love what’s coming.
...
> Friday, Nov 20
> Conductor Cloud shuts down
> We begin deleting Conductor Cloud user data. Please export your data before this date.
Oh yeah, gonna love that!
Doesn’t Claude desktop do all of this and more. Like, is this IP layer super commoditized?
T3 code is much better than conductor.
I will avoid Meta and Google as much as possible.
Condoleances
FUCK
im with you. Try free, open-source drop in replacement https://github.com/bestboyhq/malini
"Our incredible journey"
[dupe] of https://news.ycombinator.com/item?id=50369840
Why does their page sound like donald trump
Some more discussion: https://news.ycombinator.com/item?id=50369840
why sellout your company and your customers and community?
(As someone who isn't particularly happy about this news)
> why sellout your company and your customers and community?
Why not? Do you think this is a heart project, that the founders have a personal passion for agent orchestration somehow? They're not curing cancer or supporting orphans, most companies are created to make money, that's how our economic system works (for better or worse, but I'm not going to blame individuals or be surprised about it).
"community"... Come on. There's customers, there's no community here.
For money.
Like Brian Acton did with the victims he allured to his platform back in the days. Sold them for 20 billion meta coins
> The whole team is around to help answer any questions
Why did you sell out?
Realistically for them, coding harnesses are an extremely competitive landscape, and they have to live in constant fear that Anthropic/OpenAI etc will rugpull them and force these harnesses to go through API billing (effectively killing it as an option for I would guess the majority of their users). That nearly happened earlier this year when Anthropic intimated they were going to do this.
The other reality is, there are now good free alternatives to Conductor (I haven't tried them so we will see) - it does suck if you feel very tied to your existing workflow, but I also think about it from the level of 'these people gave me something for free that helped me a lot, and I want things to work out well for them'.
Why do you think they sell out? What is it about the deal that you aren’t happy about?
Users chose them knowing full well the risks.
Build a product with paid customers, then you can choose not to sell out.
Users did not know full well the risks. You never do.
And thanks for the projection. I have a product with paid customers and I refuse to sell out. The project sinks with me.
$
Are you asking people why they like money? like, a lot of money?
I'm so glad I haven't even heard of them