OpenSwarm
A swarm of agents, in the same place you work
721 followers
A swarm of agents, in the same place you work
721 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
Atlassian LoomShow, don't just tell with Loom
Promoted
being able to stop the whole swarm from one card is handy. what happens if two agents try to edit the same file at once?
@jeetendra_kumar2 So the agents can share a folder, but none of them can quietly overwrite another's work.
Before an agent edits a file it has to read it, and if someone else changed the file after that read, the edit is refused and the agent reads it again first.
For big parallel jobs, it works best to give each agent its separate projects / files
@eric_f_zeng I like the Idea of agents working together instead of having 10 different chats open. What kind of tasks work best with a swarm?
@rohit62661 Thanks Rohit! A swarm does best on work that splits into independent pieces (btw its designed more so to do write tasks, not just read tasks).
Some examples:
- filling out forms, navigating through socials / dm / messaging (all without mcp but the functionality is available if you prefer it).
- Researching 10 companies at once, one agent per company, then merging the findings into a comparison sheet
- Building an app where one agent does the page, another the data and another the tests
Step-by-step work, where each step needs the one before it, gets nothing from a swarm, so one agent is the better fit there. Splitting the job up and putting it back together is where you save
@eric_f_zeng @rohit62661 I have it run ads and script my content oh I also duped instict aswell
Very cool! What are you finding to be the best use cases?
@bullmeza So far the best ones are research that turns into a spreadsheet or live dashboard, custom tools built on demand, campaign pages in your brand voice, and jobs that cross Gmail, Calendar and Slack.
A good amount of people use this for shopping online too!
@bullmeza For me i have it script my content by scraping all the creators in my niche, running the ads for us, and simply using it as my os.
That's impressive !!!
But the question is how you manage the overall data base , is it local on the user's device or you gonna take it to the cloud ?
BTW, congrats to evry direct indirect contributor in the project. @haylli_weintraub @alex_dakhli @eric_f_zeng @benln
congrats on the launch. could you elaborate on the "where agents run the whole machine: apps create themselves, browsers control themselves, and swarms of agents work together?" perhaps?
@swati_paliwal yeah!
Apps create themselves: Describe a tool you wish you had ("a dashboard that tracks these 10 competitors") and an agent builds a working app, with source code in a folder you own. It runs as a live card on your canvas, and you can keep asking for changes.
Browsers control themselves: Agents open their own browsers on the canvas and click, type, fill forms and read pages like a person would. You watch every step and can step in at any time.
Swarms work together: Give a big job to one agent and it splits the job up, starts helpers that work in parallel, each as its own card, and brings their findings back together.
The whole machine means agents use what you use: your files, a terminal, the browser and your connected apps like Gmail, Slack and Notion. Everything stays on your computer, so you're directing the work instead of doing each step yourself.
how many levels of agents are created? agents can have sub-agents that can have sub-agents?
The shared memory across models caught my eye. Having to explain the same project again whenever you switch tools gets old pretty quickly. Congrats, Eric and team!
@oleksii_sekundant Thank you! Still making improvements all around!
@oleksii_sekundant Thanks for the support