Launched this week
Soromi
A small, fast home for AI coding agents
12 followers
A small, fast home for AI coding agents
12 followers
Most agent tools assume one folder, one login, and a window that stays open. Soromi is a workspace manager for AI coding agents: each project gets its own terminals, its own login (work, personal, others), and only the folders that matter, and you switch between them like Slack workspaces. Agents keep running when the window closes, and you can answer their prompts from your phone. Deliberately small: no boards, no personas, no dashboards. It runs the agents you already use (Claude Code, Codex).
Hey Product Hunt! 👋
I built Soromi because my daily reality with AI coding agents didn't match the tools I was using. My projects aren't one folder: there's an api repo, a web repo, a mobile repo. I have a work login, a personal login, and a friend login. And agents keep stopping mid-task.
So Soromi gives every project its own workspace, like Slack:
🖥️ Each workspace has its own terminal and agent, pointed at just the folders that matter
🔑 Each one uses its own login, so work, personal, and client accounts never mix
🔄 Agents keep running when the window closes; come back and everything is where you left it
📱 Check on an agent from your phone and answer its prompts with a tap
🔔 A notification when an agent needs you or finishes, even with the app closed
Just as important is what it will never be: no boards, no personas, no prompt libraries, no dashboards. It runs the agents you already use (Claude Code, Codex) and stays out of the way.
Being able to see a quick diff or the last few lines an agent output before responding from my phone would make remote approvals way less risky, maybe a tiny preview pane in the mobile prompt screen.
@buyuksank72065 Yes actually you will see everything, Full terminal, skills, open more agent sessions and etc all with remote control.
One thing that would seal the deal for me is per-workspace environment variable and secrets management, so each project login can pull its own API keys without me exporting them manually each time. Right now swapping contexts still means I keep a notes file of which env belongs to which workspace, and that kind of defeats the cleanliness of the whole setup. Even a simple encrypted vault tied to each workspace would do the job, no need for a full dashboard around it.
@kymetrayw You mean like agent can access to a specific environment variables per workspace ?
Finally something that gets the multi-account mess right. Switching between work and personal Claude Code sessions without logging in and out feels like the obvious thing that nobody built until now.
how does it handle the agents running in the background when the laptop is off, do they pause or do you need a server running somewhere?