How I built oncell.ai as a solo developer in one week
Hey everyone! I'm Anup, the maker of oncell.ai. Wanted to share the story behind it.
Why I built this
I run justcopy.ai, an AI coding agent with 100k users. Each user needs their own files, database, and compute — completely isolated from every other user.
Building that infrastructure took months. Docker, ECS, EFS, DynamoDB, Step Functions, security groups, IAM policies, pause/resume logic. All before I could focus on the actual agent.
I talked to other founders building AI agents. Same story every time. Everyone wires up the same AWS services from scratch.
So I extracted it into a product. Built the entire platform in one week.
What I shipped in 7 days
- Rust host agent with per-cell process isolation and cgroup resource limits
- Go control plane with scheduling, idle detection, and host registration
- TypeScript API server with auth, billing, and cell management
- Next.js dashboard with Clerk auth and Stripe prepaid billing
- SDK published on npm (@oncell/sdk)
- S3 snapshots for pause/resume with 200ms wake time
- Secrets management (env vars, never on disk)
- Per-cell observability — workflow journal, logs, metrics
- Live preview URLs for every cell
- Demo coding agent that generates web pages from natural language
- 5 CDK stacks deployed to AWS
- Docs, landing page, and this launch
How
Solo developer + Claude Code. It was like having a full engineering team - Rust, Go, TypeScript, React, CDK, all in one session.
What I'd love to know from you
- Are you building an AI agent that needs per-user compute?
- What's the most painful part of your infrastructure right now?
- Would you use something like this? What's missing?
Happy to answer any questions about the architecture, pricing, or how it works under the hood.

Replies
Be the first to reply
Have a question or a thought to share? Add a comment above to start the conversation.