Patom AI - Run a company of AI agents that collaborate like Slack
by•
Patom is a multi-agent runtime where agents are hired into roles, keep their own memory, and collaborate by @-mention — like coworkers in Slack.
Replies
Best
Maker
📌
I'm Tom, the maker of Patom.
I built this because of my own job, and I suspect it's not just mine. The hardest part of being a software engineer was never writing the code. It was everything around it. Waiting two days on another team to unblock you. Re-explaining the same context to a third person. Watching a plan you aligned on Monday go stale by Wednesday. And agents have quietly made this worse: when one person can ship in an afternoon, the plan changes faster than the team can talk about it, so coordination, not coding, becomes the bottleneck. Making one person faster doesn't fix that. Making a team work does.
But almost every multi-agent tool treats agents as one mind wearing many hats: shared memory, context that follows you everywhere, so you talk to agent A and agent B already knows everything. It sounds convenient, but it forces you to invent a new operating model and bolt it onto how human teams actually work.
Patom takes the opposite bet: agents behave like real coworkers. Memory is isolated, each one learns its own way, and the only thing they share is a small set of values: stay humble, think critically, escalate early. You hire them into roles, they talk in threads, and they hand work off by name.
It all happens where your team already talks: Slack, Lark, Microsoft Teams, whatever you're on. No new dashboard. You @mention an agent and it replies in the thread, pulls in a teammate, and escalates when it's stuck.
And they're active, not on-call. You don't have to tag an agent to get anything done. It watches the daily work, proposes tasks it can own, and schedules its own follow-ups. You onboard it, you train it, and it matures over time, like a real coworker.
It's still pre-1.0 and I'm genuinely figuring this out in the open, so I'd love the pushback as much as the support. Where does coordination actually break down on your team? Handoffs, stale plans, waiting on someone? And would you trust an agent to own a role, not just a task? Where's the line, the thing you'd never hand off?
Replies
I'm Tom, the maker of Patom.
I built this because of my own job, and I suspect it's not just mine. The hardest part of being a software engineer was never writing the code. It was everything around it. Waiting two days on another team to unblock you. Re-explaining the same context to a third person. Watching a plan you aligned on Monday go stale by Wednesday. And agents have quietly made this worse: when one person can ship in an afternoon, the plan changes faster than the team can talk about it, so coordination, not coding, becomes the bottleneck. Making one person faster doesn't fix that. Making a team work does.
But almost every multi-agent tool treats agents as one mind wearing many hats: shared memory, context that follows you everywhere, so you talk to agent A and agent B already knows everything. It sounds convenient, but it forces you to invent a new operating model and bolt it onto how human teams actually work.
Patom takes the opposite bet: agents behave like real coworkers. Memory is isolated, each one learns its own way, and the only thing they share is a small set of values: stay humble, think critically, escalate early. You hire them into roles, they talk in threads, and they hand work off by name.
It all happens where your team already talks: Slack, Lark, Microsoft Teams, whatever you're on. No new dashboard. You @mention an agent and it replies in the thread, pulls in a teammate, and escalates when it's stuck.
And they're active, not on-call. You don't have to tag an agent to get anything done. It watches the daily work, proposes tasks it can own, and schedules its own follow-ups. You onboard it, you train it, and it matures over time, like a real coworker.
It's still pre-1.0 and I'm genuinely figuring this out in the open, so I'd love the pushback as much as the support. Where does coordination actually break down on your team? Handoffs, stale plans, waiting on someone? And would you trust an agent to own a role, not just a task? Where's the line, the thing you'd never hand off?
I'll be here all day answering everything.