You can now close your laptop and your agents will be still running
by•
Hey Product Hunt!
Murmell is now live!
We built Murmell around a simple frustration: AI coding agents are powerful, but most of them stop working as soon as you close your laptop. Murmell gives your agents a shared workspace in the cloud. You can launch several agents on the same repository, watch their terminals, coordinate their work, and come back later without keeping your computer awake.
We would love your feedback, especially on one question: what do you currently do when you need your coding agents to keep working after your laptop is closed ?
201 views
Replies
I've been using a spare machine for long running agents simple but not ideal
Murmell
@hambali_salisu Switch to the future ! Switch to Murmell ahaha !
Murmell
@hambali_salisu Btw we are live in the main age, let's support us :)
Running everything on a cloud VM and reconnecting when needed has worked pretty well for me.
Murmell
@_tina_kim
Yeah, that’s pretty much the approach we wanted to make simpler. The main difference with Murmell is that you can have multiple agents running in the same workspace and just reconnect when you’re back. No need to keep a VM setup or your laptop running.
Murmell
Let's win the day, let's support Murmell guys fo the win !!!!!
Keeping the laptop awake works but the battery drain gets pretty fast lol.
Murmell
@_logan_myers Yeah haha, that’s definitely one way to do it 😅 but having your laptop running just to keep an agent alive feels a bit wasteful. That’s exactly the problem we wanted to solve with Murmell!
Murmell
@_logan_myers Btw we are live in the main age, let's support us :)
MonoCloud for Startups
The shared workspace approach is interesting. I’ve seen other multi-agent setups isolate each agent into its own worktree/branch and deal with coordination later at the Git/PR layer. You’re doing almost the opposite with multiple agents working against the same project and using The Board to reserve files before they touch them.
Was that a deliberate choice from the beginning? I’m curious whether you found shared context between agents more valuable than the safety you get from isolating their work.
Murmell
@roguetink Yeah, it was pretty deliberate. We found that shared context can be much more valuable when you have multiple agents working on the same project.
With isolated worktrees, you get a nice safety layer, but you also add a lot of coordination overhead between agents. We wanted to experiment with the opposite approach: let agents see the same project and coordinate closer to where the work actually happens.
Still figuring out the best balance though 👀
I currently rely on GitHub codespaces and just leave the tab open, which feels hacky and wastes resources. Does Murmell let you set priorities so one agent waits while another finishes a shared file?
Murmell
@new_user___209202627e87af67bf41b28 Yeah, I feel you becauseleaving a Codespace tab open is definitely a bit hacky 😅
For the priorities part, not quite in that exact way yet. Agents currently share the same workspace and can see what the others are doing, but explicit file-level task scheduling is something we're working towards.
That's actually a really interesting use case though 👀
Murmell
@new_user___209202627e87af67bf41b28 Btw we are live in the main age, let's support us :)
Keeping a laptop awake just so an agent can finish a long task is a surprisingly awkward limitation. The shared cloud workspace sounds useful, especially when several agents are working on one repository. How does Murmell handle conflicts if two agents edit the same files at the same time?
Murmell
@sampurna_seo Yeah, this is one of the interesting problems with multiple agents sharing a workspace.
Right now, agents can see the same project and what the others are working on, but we’re still improving the conflict-handling side.
The goal is to make it much harder for two agents to step on each other while still keeping the benefits of a shared workspace.
Dial
Right now it's the same "leave the laptop open" approach others mentioned, which mostly works until the agent hits something it needs a decision on. That's the part I'm curious about with Murmell: when an agent is running unattended in the shared workspace and hits a permission prompt or an ambiguous choice, does it just block and wait silently until someone reconnects, or is there a notification that pulls you back in? A cloud workspace solves the "keeps running" half, but the "still needs you occasionally" half seems like the harder problem.
Murmell
@galdayan As I said in other comments, when the agent needs you, it goes into what we call "needs-you" mode. Its window starts blinking on the workspace map, and its status in the agent bar on the right blinks too, so you don't have to babysit every tab to notice.
Right now that's the visual signal; we're planning to layer a proper notification system on top, mobile app push, that kind of thing so you get pulled back in even when you're not looking at the screen.
I'm curious to hear from you: if we built that, is it something you'd actually use?
Dial
@moss_ab_mirande_ney yes, I'd use it, but I'd push for piping it into a channel people already check rather than building a new mobile app from scratch - SMS or a phone call for the truly blocking cases, Slack/email for the lower-urgency ones. a new app is one more thing to install and grant permissions to, and "needs-you" moments are exactly the kind of urgent-but-rare event people miss if it's siloed in a tool they only open when they remember to. the actionability point Anuj raised is the harder half though - a notification that just says "blocked" without the specific question isn't much better than the blinking tab.
Murmell
@galdayan it's a very good point, and i think it's way better the channel pushing thing, and also in terms of development time, it's way less complicated than a complete separate app. And do you see it juste like a notification telling you that this agent or this agent is blocked or something more detailled about what your agent specifically need you to do ?
@galdayan same wall i hit further up this thread. from outside, an agent that blocked and an agent that guessed wrong look identical until you go and read the trace. so the version i want is one where it has to say what it was unsure about.
I usually stop my agents before closing my laptop because I do not trust long running tasks unattended. Could Murmell add automatic checkpoints so I can review what changed before merging?
Murmell
@noahanderson Yes ! it's up to you and your CI CD, the agent is waiting for you to do the critical tasks,
and we are thinking to add a special status soon for all the github operation (something like ready to merge etc)
Would this be a feature that you find useful ?