Launching today
OpenSwarm
A swarm of agents, in the same place you work
609 followers
A swarm of agents, in the same place you work
609 followers
Stop bolting AI onto your computer. OpenSwarm is an AI-first operating system where agents run the whole machine: apps create themselves, browsers control themselves, and swarms of agents work together. Your OS just became an infinite canvas, and your work just transformed from you doing one task at a time, to orchestrating dozens in parallel. Tell it what you want and a swarm of agents does the work with you at the helm.
Free
Launch Team / Built With
Viktor.comAn AI coworker that actually does the work
Promoted
Hey Product Hunt 👋 I’m Eric, co-founder and CTO of OpenSwarm.
Your AI tools can do a lot. But you can still end up connecting everything: moving files between apps, repeating the same context, carrying one model’s answer into another conversation, and tracking what each agent is doing.
You shouldn’t have to be the shared memory between your AI tools.
We’ve spent the last four years at Berkeley researching how agents work together: how they share memory, coordinate tasks, use browsers, and build on each other’s work. OpenSwarm brings that research into a product designed for everyday people, not just developers.
Our belief is simple: making models smarter is only part of the problem. We also need to change the environment they work in.
That’s why we built OpenSwarm: an AI-first operating system designed for you and a swarm of agents to work together. An army of agents, in the same place you work.
It runs on your own computer, across desktop platforms, bringing your models, apps, files and agents into one infinite canvas.
Here’s what that means in practice:
🧠Shared context across models and conversations. The memory, skills and workflows you build in OpenSwarm aren’t tied to one model or one chat. Switch models without rebuilding your saved context. Agents can share findings and build on work that’s already been done.
🔗Your existing tools, connected. Bring together Gmail, Calendar, Drive, Slack, Notion, GitHub and more, alongside your local files and the browser accounts you’re already signed into. Give agents a job that crosses tools, rather than moving the information between them yourself.
🐝Agents working together, not just more chats to manage. Send a swarm to research different sources, analyze files or build software in parallel. Their browsers, apps and documents appear as live cards on the canvas, where you can inspect, interact with and guide the work. Outward-facing actions, like sending emails or publishing posts, wait for your approval.
For example, ask OpenSwarm to research a set of companies using the web and your project files, turn the findings into a comparison spreadsheet, and build a dashboard to track them. The agents can divide up the research, share what they find and create the outputs in the same workspace.
And when you need a tool that doesn’t exist, describe it. OpenSwarm can build a working app with source code in a folder you own, running alongside the rest of your work.
The shift we’re building for is from doing every task yourself to directing work happening in parallel—without losing the context or control.
OpenSwarm is free, and access is currently waitlist-only. You bring your existing Claude or ChatGPT subscription, or your own model credits; those costs are separate from the free OpenSwarm app.
Join the waitlist on our website to request access. We’d love your feedback as we build this.
What’s the first job you’d give your swarm?
how many levels of agents are created? agents can have sub-agents that can have sub-agents?
Awesome :D
@raezil Thanks!
@raezil Thanks you!
Love it!
@francois_levy thanks!
@francois_levy Thankyou soooo much
I just generated a page for our campaign and it pulled our brand voice in like two mints. pretty solid execution.
@charles_eric3 Sweet.
Making the OS itself an infinite canvas where agents can create apps and run the machine is a bold direction, and moving from doing one task to orchestrating many is the right way to frame it. Swarms of agents working together is the part worth watching. How do you keep a swarm from stepping on itself when several agents touch the same surface?
@karimbenkeroum yeah, of course. But I don't see chatbots being the main way you work with many agents at once.
The bigger idea is collaboration.
Everything a job leaves behind (memory, apps, workflows, skills) stays in your workspace, so each new agent builds on the last and they keep getting better.
You can also pack a whole board into one .swarm file, with your keys and logins stripped out, and hand it to a teammate to pick up where you left off. The more people and agents build on it, the more every new agent starts with. That network effect is what we're building toward.
the file conflict handling (read, check if changed, refuse and re-read) makes sense for shared files, but what about actions that aren't a file write, two agents both sending the same DM, both submitting the same form, both hitting an API that isn't idempotent. there's no version number to check against on those, so what stops two agents from quietly duplicating an outward action instead of a file edit