Apr 25, 2026
Unideploy on Verges
# UniDeploy — Building Vercel, But Self-Hosted and Extensible
> Think Vercel, but you own the infrastructure.
## The Problem
Every solo developer or small team hits the same wall eventually.
You've written good code. You have a GitHub repo. But getting that code running reliably on *your* infrastructure — with proper builds, rollbacks, observability, and zero downtime — suddenly feels like a second job.
Managed platforms like Vercel or Render solve this, but at a cost: you give up control. You can't customize the runtime, you're locked into their pricing, and the moment you need something non-standard, you're stuck.
I wanted something different. A deployment platform that **abstracts the complexity of AWS, Kubernetes, and Docker** — but runs entirely on infrastructure you own.
That's why I built **UniDeploy**.
---
## What UniDeploy Does
UniDeploy is a one-click deployment platform designed for developers who want Vercel-like simplicity without giving up infrastructure control.
You push code. UniDeploy handles the rest — building, deploying, and notifying — all on your own AWS EKS cluster.
---
## The Architecture: Multi-Agent System
The most interesting engineering decision in UniDeploy is the **multi-agent architecture**. Instead of one monolithic deployment process, three specialized agents handle the workflow:
### 🔨 Build Agent
Takes your source code, runs the build pipeline, creates a Docker image, and pushes it to the container registry. It's isolated — if the build fails, nothing downstream is affected.
### 🚀 Deploy Agent
Pulls the built image and orchestrates the rollout on Kubernetes. It handles rolling updates, health checks, and rollback logic if the new version doesn't pass readiness checks.
### 🔔 Notify Agent
Once deployment succeeds (or fails), it fires notifications. You always know exactly what happened and when.
This separation of concerns makes the system **composable and debuggable**. When something goes wrong, you know exactly which agent failed and why.
---
## Real-Time Observability with WebSockets
One thing I always hated about deployment tools: you click deploy, and then... you wait. And you don't really know what's happening.
UniDeploy uses **WebSocket connections** to stream real-time logs back to the dashboard as your deployment progresses. You see the build output, the Kubernetes rollout events, and the final status — live, as it happens.
No more refreshing. No more guessing.
---
## The Stack
| Layer | Technology |
|---|---|
| Orchestration | AWS EKS (Elastic Kubernetes Service) |
| Containers | Docker |
| Real-time Logs | WebSockets |
| Agent Communication | Multi-agent message passing |
| Infrastructure | AWS (self-hosted) |
---
## Why Kubernetes?
K8s gets a reputation for being overkill. For some use cases, it is.
But if you want production-grade deployments with:
- **Rolling updates** (zero downtime)
- **Horizontal scaling**
- **Health checks and self-healing**
- **Namespace isolation** between projects
...Kubernetes is the right foundation. UniDeploy abstracts it so you don't have to write YAML by hand for every deploy — but it's all K8s underneath, which means you can inspect, extend, and customize it fully.
---
## What I Learned Building This
### Multi-agent systems need clear contracts
The biggest design challenge was defining exactly what each agent is responsible for — and what it is *not* responsible for. Blurry boundaries cause bugs that are hard to trace. Each agent in UniDeploy has a strict input/output contract.
### Observability is a feature, not an afterthought
Real-time logs via WebSockets weren't a "nice to have." They're the difference between a tool you trust and one you're afraid to use.
### Self-hosted doesn't mean unsupported
One of the goals with UniDeploy was to make self-hosting feel as smooth as a managed platform. That means good defaults, clear error messages, and documentation that actually explains *why*, not just *what*.
---
## What's Next
UniDeploy is actively being developed. Planned next steps include:
- **GitHub Actions integration** — trigger deployments automatically on push
- **Multi-environment support** — separate staging and production pipelines
- **Plugin system** — extend build and deploy steps without forking the core
- **Web dashboard** — visual deployment history, rollback UI, and environment variable management
---
## Try It
**GitHub:** [github.com/CoderRahul01/unideploy](https://github.com/CoderRahul01/unideploy)
If you're tired of paying for infrastructure you don't control, or you want to understand how deployment platforms actually work under the hood — UniDeploy is worth a look.
---
*Built by Rahul Pandey — developer, builder, DevOps tinkerer.*
*Twitter/X: [@rahulpandey187](https://x.com/rahulpandey187) · GitHub: [CoderRahul01](https://github.com/CoderRahul01)*
Read Entry →
