Murmell – Collaborative cloud canvas for coding agents
Hey HN! I'm Moss'Ab. Murmell (https://murmell.com) is an infinite canvas where coding agents run together in the cloud instead of on your laptop. It's like Google Docs, except the other editors are you, your teammates, and a set of agents all working in the same canvas. Short demo: https://app.murmell.com/demo The idea came out of a YC hackathon. We were trying to build the thing under extrem pressure, deploying at the very last minute, but with several agents working in parallel, all the branches collided. Everything broke at the worst possible moment and we spent…
What it does
Collaborate with parallel AI coding agents in a shared cloud sandbox. Prevent merge conflicts, lock file paths in real time, and deploy from any terminal.
Share one canvas instead of passing screenshots around. Everyone sees the same terminals, cursors and preview while the work is happening. Open Claude Code, Codex, Kimi and more in the same project. Move, resize and follow each session without losing the thread. When two agents reach the same file, Murmell lets one edit while the other waits. No overwritten work, no manual coordination. Choose the live canvas now. Add Murmell Bots when they arrive. Murmell is a browser canvas where several AI coding agents (Claude Code, Codex, Kimi, OpenCode) run on a cloud machine and share one project directory. Each agent opens in its own window with a real terminal, and everyone on the link sees the…from murmell.com
In the maker’s words, at launch
Hey HN! I'm Moss'Ab. Murmell (https://murmell.com) is an infinite canvas where coding agents run together in the cloud instead of on your laptop. It's like Google Docs, except the other editors are you, your teammates, and a set of agents all working in the same canvas. Short demo: https://app.murmell.com/demo The idea came out of a YC hackathon. We were trying to build the thing under extrem pressure, deploying at the very last minute, but with several agents working in parallel, all the branches collided. Everything broke at the worst possible moment and we spent the end of it untangling instead of building. We didn't win. Murmell is the thing I wished we'd had that day: one place where the agents and the people can see each other work, on machines that don't belong to any one laptop. Each canvas gets its own machine in the cloud. Close your laptop and the agents keep going; open it again anywhere, or have a teammate open the same canvas by sending him the link, and you're inside the same session rather than a copy of it. The work stays exactly where you left it : same branches, same dev server still running, so an agent picks up mid-task instead of starting from scratch. And you can literally collaborate with your team like google docs, it's se same system, read of edit link, they connect they can type on your terminal, you can see their mouse, wich terminal they're typing in. And you can interact wich each other sessions, it's pretty useful instead of sharing a screen, copying a prompt, etc. The hardest part was building the cloud infrastructure, trying to make it scalable, and to manage the sessions to make the best experience possible: when the VM shuts down, it snapshots everything and restores everything when you come back, even the conversation with your terminal, it's stored with a system that i built like obsidian wich also make you save a little bit of tokens. So the answer to that hackathon is file claiming. Before an agent works on something, it claims the paths, it's an exclusive lease with a TTL, so nothing stays locked forever because an agent died mid-task. And if another agent asking for the same file gets denied, the denial tells it who holds it, how long is left, and to either pick other work in its scope or message the holder (the agents can communicate through the board) A watcher classifies every write, so if anything writes inside someone else's claim it shows up on the canvas as a collision right then, instead of as a merge conflict you discover an hour later. And when an agent is done, it can hand its claim to the next one. And agents never hold your provider keys either, they get a token, and a proxy on the container's loopback swaps it for the real key on the way out, which makes a leaked token worthless outside that container. Murmell is paid. I've put prices at 50% off for now, because it's still not where I want to take it yet. It is fully usable though: I develop Murmell with Murmell, every day (and tbh it's save soooo much time to us with my cofounder) And to be fully transparent with the actual prices it costs us much than we are earning but the goal is to collect feedback to make Murmell as good as possible For context: I'm 19, a 4th year computer engineering student, and I've been on this for about three and a half months, day and night. I discovered claude code after my father showed me first Windsurf (today Devin) for a python homework in 2024, and i think i never went somewhere without my laptop since then lol. So feel free to comment or to give feedback, or even subscribe to support the project :)
Does the same job
all alternatives →
More dev tools this month
the category →
Open-source GTM skills for technical founders
Dev tools · 29d ago · gtmcofounder.com
OpenTrailPaper is open-source bike computer firmware for the LilyGO T5S3 4.7" E-Paper PRO. It supports offline maps, GPX routes, FIT recording and Bluetooth sensors.
Dev tools · 1d ago · opentrailpaper.com
Launched alongside, August 2026
the whole month →- TL
Life & fun · 10d ago · louisabraham.github.io
- SA
Hello HN! I found that picking out plausible but diverse skin tones for my digital art and game development projects was kind of difficult, and I got curious about if there was a way to define a color space that made it easy. I've built a color picker and procedural generation algorithm based on the space as well as a bunch of other fun js features and demos throughout the page that use the equations. If you find it interesting, I have lots of explanations of how I built it and what properties the space has. The methodology might be a bit shaky, but hopefully the result is as helpful for…
Life & fun · Aug 2026 · toneyalexander.github.io
I trained a 125M-parameter transformer to autocomplete piano performances in real time (~108 notes/sec on an iPhone 15). The idea is basically GitHub Copilot or Tabnine, except instead of prompting it with code, you prompt it by playing a few notes on a MIDI piano. The model then continues what you played, entirely on-device. The app is free if anyone wants to try it. Happy to answer questions about the model, training, Core ML, or the many things that didn't work.
AI · 17d ago · simedw.com