redu.cloud
Run machines at agent speed, not human speed.
75 followers
Run machines at agent speed, not human speed.
75 followers
redu.cloud is a cloud made for AI agents, with a generous free tier. It allows your agents to get the stateful resources they need extremely fast, making them ideal for doing deployments of your app, DevOps, training stateful AI models and fixing live incidents at agent speed, safely.
Free Options
Launch Team / Built With
CloudflareYour next app belongs on Cloudflare Workers. Deploy for free
Promoted
machines at agent speed is a fun framing 🚀 how fast is cold start actually?
@petrkovacikDepends on the flavor size. For small ones it is ~ a few seconds. But we are not provisioning sandboxes. We are provisioning an entire heavy cloud for an agent: machines, network, storage... All stateful and can run for days/weeks/years. And all is provisioned at the intent stage, so in most cases the agent gets what it needs before the “thinking” stage is over.
MCP-only is a bold choice and I think it's the right one — bolting an agent interface onto a human dashboard usually leaves agents as second-class users. What I'm wondering is the human side of it: when an agent is provisioning and tearing down machines on its own, what do I get to see and control? Logs of what it did, spending caps, a big red stop button? That visibility is what would make me comfortable letting it run unsupervised.
@jcbuilds This is a great question. On one hand we give agents full control of resources, but resources themselves track everything agents do. This also solves the problem of multiple agents working on the same task. On the human side of things you can always see logs (also in the dashboard), and there is a spending cap you can define per account. You can delete any resource via the dashboard, or access them manually (with SSH for example). You still have full control.
Hey Product Hunt 👋
I'm Miloš, co-founder of redu.cloud with Aleksandar.
I have worked with infrastructure for the last 7 years. Agents came. I noticed that I was spending half of my day just watching and waiting for Claude Code to finish infra work.
So we sat down and reconstructed the entire cloud orchestration and its parts to fix this.
It turns out that if your cloud works almost instantly for agents, fully autonomous agentic DevOps becomes reliable, large stateful ML training can be done in hours, coding and deployment become (even) quicker, and an entire ocean of agent-cloud opportunities becomes possible.
Come and grab our freebies (1 VM, 1 Postgres, 100% free, MCP only).
Tell us the regular or crazy things you want to try to run on it! We're here all day.
Fast provisioning makes cleanup correctness more important, not less. I’d want every agent-created resource tagged with run ID, owner, budget ceiling, and expiry, plus a receipt showing what was actually destroyed. That turns autonomous infrastructure from “easy to create” into something operators can safely reconcile.
@keesan12 Thanks a lot for the feedback. We already have clear budget ceilings and a trail of everything done, destroyed, created and so on, but not in the form you described. We can add that.
Machines that spin up stateful resources at the speed an agent works rather than the speed a person waits for is the bottleneck I keep hitting, and a free tier makes trying it a no brainer.
@karimbenkeroum That's great to here, i look forward to hearing about your experience with the platform :)