®
Circloud
NEST
®
Circloud
NEST
Local Cloud Platform

Self-hosted Cloud Platform

Run Circloud® NEST on a cluster of Linux servers or as a virtual cluster on a Desktop Computer, and get a full private cloud on hardware you own: VPCs (Virtual Private Clouds) with their subnets and pods, persistent and S3 cloud storage, DB clusters and email, all driven by one REST API.

Every app gets a public URL on your own domain through a free Cloudflare tunnel, with the edge’s DDoS protection and TLS in front of it. One click installs the cluster, your AI agents deploy to it through the API, and it grows server by server, up to 100 servers per cluster on the largest plan.

No public IP. No open ports. No hyperscale bill.

HOME HARDWARE
Cluster of Servers
or
Virtual Cluster on Desktop Computer
No Public IP
Outbound · straight to the internet, bypasses the edge
EDGE
Cloudflare Tunnels · FREE
Public URL
Public URL
Public URL
INTERNET
USERS
Public URL
Your hardware → Cloudflare edge → your users

Key features

Networking

  • Software-Defined Networking
  • Public URLs & Custom Domains
  • Security Rules & Isolation

Compute

  • Auto-Healing & Auto-Scaling
  • GPU Access
  • Secure Console Access

Data & Services

  • Persistent Storage & S3 Cloud Storage
  • DB Clusters
  • Email

Operations

  • One-Click Installer
  • Infrastructure Scaling
  • VPC Migration
  • Monitoring & Alerts

Interfaces

  • REST API
  • Rich Graphical User Interface
See it in action

Built for

Vibe coders & Web App Pre‑Ventures

An app developer writing code at a home-office desk, a tower PC and a router beside him
MIST™

Take your apps live from the desktop you code on, with no hosting bill while your idea takes shape.

  • 1 Desktop Computer
  • 1 virtual cluster (up to 3 virtual nodes)

Solopreneurs

A solopreneur coding at her desk, three rack servers and a router on a low rack beside it
CIRRUS™

Spend your time on your business, not on hosting: run your products and your clients’ sites on servers you own.

  • Up to 3 servers per cluster
  • 2 clusters (up to 6 servers)

Co‑founded startups

Two co-founders coding side by side in a garage office, their server rack against the wall
NIMBUS™

Give your founding team its own cloud from day one, and spend your runway on the product.

  • Up to 8 servers per cluster
  • 2 clusters (up to 16 servers)

Mid‑size businesses

An open-plan office of a mid-size business, its three server racks behind a glass wall
ANVIL™

Turn the servers your company already owns into a private cloud, and stop renting what you already have.

  • Up to 25 servers per cluster
  • 2 clusters (up to 50 servers)

Datacenters

A datacenter aisle lined on both sides with numbered racks of servers
SUPERCELL™

Put your own cloud on the racks you operate, and give your clients more than rack space.

  • Up to 100 servers per cluster
  • 2 clusters (up to 200 servers)

Production-grade on your own desktop or servers

From your app to a public URL Left to right: your app is deployed into your cluster, shop a step at a time under a progress bar (its VPC, its subnet, its pods, its tunnel), then the games topology at once under a second one. Inside the cluster two VPCs (Virtual Private Clouds), shop and games, each hold a subnet with three pods: a database, a back-end and a front-end. The front-end pod of each VPC opens its own outbound tunnel to the Cloudflare edge, and beyond the edge a public web address points back down each tunnel, so a visitor's request comes in from the edge with no VPN, no open port and no public IP. Under the edge a second path leaves the games VPC's back-end straight for the internet: outbound traffic from your pods bypasses the edge. Under the two VPCs, inside the cluster, a bar names the platform that runs them, Circloud NEST, and under it stand four servers of your own, the machines the cluster runs on. DEPLOYING SHOP GAMES TOPOLOGY creating VPC creating subnet creating pods opening tunnel shop is live creating topology games is live my-app my-app:1.0 Your App DEPLOY YOUR SERVERS ·your cluster VPC(Virtual Private Cloud) ·shop SUBNET 10.10.1.0/24 shop-db shop-api BACK-END shop-web FRONT-END VPC(Virtual Private Cloud) ·games SUBNET 10.20.1.0/24 games-db games-worker BACK-END games-web FRONT-END VPC · SUBNET · POD — a private network per app
PODS · NETWORKS · TUNNELS srv-01 srv-02 srv-03 srv-04 CLOUDFLARE https://shop.yourdomain.com LIVE https://games.yourdomain.com LIVE INTERNET OUTBOUND BYPASSES THE EDGE
Your app → your pod → your tunnel → the edge → a public URL
Swipe the diagram sideways to see all of it

Zero open ports, globally accessible

No public IP needed. Your pods are reachable through your own domain name over a secure FREE Cloudflare tunnel — only the protocols and ports you explicitly allow ever reach your network.

You don’t install a firewall, configure NAT, or open inbound ports. Cloudflare’s edge sits in front of every request and provides DDoS protection, WAF, bot mitigation, TLS termination, and rate limiting for free, before traffic ever touches your machine.

Tunnel creation is built right into Circloud — no DNS records, no cloudflared installs, no clicking through the Cloudflare dashboard. Add your Cloudflare account ID and API token on your Circloud profile once and every pod you deploy can be exposed on a custom domain in seconds.

Full cloud stack — from one machine to a 100-server cluster

Deploy VPCs, containers, and isolated networking on a single desktop or server. Start on one machine and grow server by server. A virtual cluster on one Desktop Computer (64 GB RAM) already carries 2.5K concurrent users and 125K accounts; one Xeon 5 server (256 GB RAM) carries 10K concurrent users and 500K accounts; and on the largest plan you can keep adding servers, up to 100 in one cluster — your own private cloud, on hardware you own.

You get the full cloud feature set out of the box: auto-healing pods, horizontal auto-scaling, persistent volumes, unlimited Cloudflare tunnels, monitoring, centralized logs, load balancers, NAT gateways, custom domains — every primitive you’d expect from a hyperscaler, running on hardware you own.

Capacity and Scaling
  1. 1 desktop
    Virtual cluster
    64 GB RAM
    Concurrent users 2.5K
    Daily active users 12.5K
    Accounts 125K
  2. 3 servers
    Xeon 5
    256 GB RAM each
    Concurrent users 30K
    Daily active users 150K
    Accounts 1.5M
  3. 25 servers
    Xeon 5
    256 GB RAM each
    Concurrent users 250K
    Daily active users 1.25M
    Accounts 12.5M
  4. 100 servers
    Xeon 5 cluster
    256 GB RAM each
    Concurrent users 1M
    Daily active users 5M
    Accounts 50M

Figures are per cluster. One Xeon 5 server (256 GB RAM) carries 10K concurrent users, 50K daily active users and 500K accounts. Capacity follows RAM: a virtual cluster on one Desktop Computer (64 GB RAM) carries 2.5K concurrent users, and a cluster adds up server by server, to 100 servers on the largest plan.

AI-Agents and API-First Deployment

Create and Deploy your apps with AI agents

Complete REST API

Every operation — VPCs, subnets, pods, replicas, auto-scaling, persistent volumes, Cloudflare tunnels, custom domains — available as simple HTTP calls. The same API powers both GUIs; anything you click, a curl command can do.

Built for AI agents

A natural fit for Claude Code, Cursor, Copilot Workspace, or any LLM that can issue HTTP requests. Drop your .credentials file into the project: the agent reads the whole API reference from your cluster, builds the image, then calls the API to provision VPC + pods + tunnels + live URL — in under 30 seconds, zero human intervention.

An agent builds the whole environment On the left, your own machine: an editor, a project folder holding a single .credentials file (the API's address, the tenant id and the token, downloaded from the portal), and the AI agents you already use, which read the API reference from the cluster itself; on the right, your server cluster, four servers with Circloud NEST on them, over the Cloudflare edge. A progress bar over the agents, AI agent writing code: the code falls from them into the same folder, and the app's files land in it. A second bar over the API arrow, the agent deploying: POST /vpcs and the VPC shop, POST /subnets and its subnets, POST /pods and its pods, POST /tunnels and its tunnel to the edge and its live public address. Then the single POST /topology call and a third bar, creating the games topology, and the whole games VPC at once: its subnets, pods, tunnel and address. YOUR IDE VS CODE · CURSOR my-app — □ ✕ ← → ↑ This PC › my-app Name Type .credentials Credentials AI AGENT WRITING CODE DONE AGENTS <div> const fetch() .css { … } </> my-app — □ ✕ ← → ↑ This PC › my-app Name Type .credentials Credentials src Folder assets Folder styles Folder <> index.html HTML
AGENT DEPLOYS TO
 creating VPC · shop creating subnets creating pods opening tunnel shop is live POST /vpcs POST /subnets POST /pods POST /tunnels OR POST /topology CREATING GAMES TOPOLOGY DONE THE AGENT CALLS THE API YOUR SERVER CLUSTER srv-01 srv-02 srv-03 srv-04
REST API · ONE TOKEN 200 OK VPC · shop subnet · web web api worker subnet · data db cache jobs VPC · games subnet · web web api worker subnet · data db cache jobs tunnel tunnel TLS AT THE EDGE CLOUDFLARE https://shop.yourdomain.comLIVE https://games.yourdomain.comLIVE
.credentials → code → VPCs · subnets · pods → tunnel → live URL
Swipe the diagram sideways to see all of it

Install and Scale

Simple one-click installation

Tell the installer where the cluster runs — a virtual cluster on one Desktop Computer, or your physical servers — and press Install once. It prepares every machine, joins them into one cluster, and hands back a running private cloud with its web console.

The installer itself runs on your own Windows computer, and you never set the machines up one by one.

One click builds the whole cluster An installer running on a Windows computer shows one cluster, six machines detected, and a single Install button with the mouse resting on it. Its progress bar is full and reads cluster installed. Two routes leave the installer: a virtual cluster of three nodes on a Desktop Computer, standing on the desktop's virtual switch and its virtualization layer, OR physical servers of your own — an install is one or the other. All six machines are online. INSTALLER MACHINE YOUR MACHINES Windows
Installer
 Cluster nest-01 Machines 6 detected Install CLUSTER INSTALLED CASE 1 — VIRTUAL CLUSTER DESKTOP COMPUTER VIRTUAL SWITCH WINDOWS VIRTUALIZATION LAYER node-1 node-2 node-3 OR CASE 2 — PHYSICAL SERVERS srv-01 srv-02 srv-03
One click · virtual or physical · from Windows
Swipe the diagram sideways to see all of it
Adding and removing servers in a running cluster The installer, running on a Windows computer, has an Add server and a Remove server button. A cluster of four servers, each running three pods, gets two more: they join, and the cluster's capacity grows from 32 vCPU and 128 GB of memory to 48 vCPU and 192 GB. Four pods then move onto the new servers, as a VPC's Migrate internal moves its pods to the servers chosen for it. Then the two are removed: their pods move back to the other four first, they leave, and the capacity shrinks again. With no rebuild, and the cluster running all along. INSTALLER MACHINE Windows SERVERS +Add server −Remove server 2 SERVERS JOINED 2 SERVERS REMOVED the installer you already have YOUR CLUSTER 4 SERVERS 6 SERVERS srv-01 ONLINE srv-02 ONLINE srv-03 ONLINE srv-04 ONLINE srv-05 ONLINE DRAINING srv-06 ONLINE DRAINING CLUSTER CAPACITY 32 vCPU · 128 GB RAM 48 vCPU · 192 GB RAM Add or remove servers · no rebuild · the cluster you already have 4 servers → 6 servers · no rebuild, no downtime · the cluster just gets bigger 6 servers → 4 servers · pods move first · the cluster just gets smaller
Swipe the diagram sideways to see all of it

Add a server. The cluster grows.

Outgrown the hardware? Add the new machines with the installer's Extend. They join the cluster you already have, and their capacity is available to every VPC and pod straight away.

On the largest plan the same step repeats as far as a hundred servers in one cluster — no reinstall, no downtime, and the apps already running keep their addresses and their public URLs.

Need fewer? The installer's Shrink removes a server: its pods move to the other servers first, then it leaves the cluster.

High Availability, Auto-scaling, and Auto-healing

Every pod can run multiple replicas, spread across your machines, with a built-in load balancer distributing traffic between them — one service address in front, healthy replicas behind it. Lose a machine, and the surviving replicas keep answering.

Auto-healing watches every replica. If one crashes or stops responding, the platform detects it and starts a fresh replica automatically — no pager, no manual restart.

Auto-scaling watches CPU load. When traffic climbs it adds replicas of your pod, up to your maximum; when load drops it scales back down to save resources. Both are a single switch, at creation or any time after.

Auto-healing and auto-scaling Two panels. Auto-healing: three replica tiles; one crashes and turns red, the platform restarts it, and the readout returns to 3 of 3 healthy. Auto-scaling: as users arrive the CPU meter climbs past the 70 percent threshold, extra replicas of the pod shop-web appear, and when load drops the extras are removed again, back down to the minimum of two. AUTO-HEALING ✕ restarted 3 / 3 HEALTHY 2 / 3 HEALTHY · RESTARTING AUTO-SCALING USERS × 3 × 8 CPU 70% 41% 88% 49% pod shop-web REPLICAS shop-web-1 shop-web-2 shop-web-3 shop-web-4 replicas: 2 · min 2 · max 10 replicas: 4 · min 2 · max 10
Auto-healing · Auto-scaling — your pods look after themselves
Swipe the diagram sideways to see all of it

Virtual Networks: Software-Defined Networking

Tenant-A Tenant
shop VPC
Subnetweb10.0.1.0/24
shop-web
shop-api
shop-cache
Subnetdata10.0.2.0/24
shop-db
games VPC
Subnetweb10.0.1.0/24
games-web
games-api
Subnetdata10.0.2.0/24
games-db
Tenant-B Tenant
blog VPC
Subnetweb10.0.1.0/24
blog-web
blog-cache
Subnetdata10.0.2.0/24
blog-db
analytics VPC
Subnetjobs10.0.1.0/24
etl
Subnetdata10.0.2.0/24
warehouse
SOFTWARE-DEFINED NETWORKING
srv-01
srv-02
srv-03
srv-04
srv-05
srv-06
Tenant VPC Subnet Pod Server

Layered, software-defined isolation

Each tenant gets fully isolated VPCs, carved into private Subnets, where Pods run side-by-side — all routed, addressed, and firewalled in software.

Because each VPC is a separate L3 boundary, two tenants can reuse the same CIDR ranges without ever colliding. Subnets stay private by default; outbound traffic flows through the per-VPC NAT; inbound is exposed only through Cloudflare tunnels.

Cost Comparison

Platform software licensing

Enterprise container platforms like OpenShift, Tanzu, or Rancher charge per-core or per-node licence fees that can reach tens of thousands of dollars per year — and those fees renew annually.

Circloud® NEST is a one-time purchase of $11.85, with no recurring licence fees, no per-core charges, and no vendor lock-in.

Platform licensing — annual cost (3-node cluster)
Red Hat OpenShift $13,000 – $30,000+ / yr
VMware Tanzu $5,000 – $20,000+ / yr
SUSE Rancher Prime $7,300 – $25,600 / yr per node
Replit $25 – $100 / mo ($300 – $1,200 / yr)
$11.85 one-time
One-time purchase — no recurring licence fees, no per-core charges
$13k+
OpenShift / yr
per 3-node cluster
$11.85
— one-time
no recurring licence fees

Infrastructure hosting

Running 4 production instances (4 vCPU, 8 GB each) with managed containers, a database, load balancer, and storage costs $400–$800+ per month on AWS, Azure, or GCP.

With Circloud® NEST, the same workloads run on hardware you already own — your electricity bill is the only recurring cost.

AWS / Azure / GCP
$0 $500 $1,000
$800+ / mo
4× m5.xlarge + EKS + RDS + ALB + storage + egress
Your Desktop/Servers
$0 $500 $1,000
$0 / mo
no recurring fees — hardware you own
~15x
Lower hosting cost
$0
Recurring fees

Feature Comparison

How Circloud® NEST compares to managed cloud services and enterprise container platforms across features, pricing, and operational overhead.

Feature Cloud Providers
AWS / Azure / GCP		 Enterprise
OpenShift / Tanzu / Rancher		 Replit
Price $11.85 one-time $400 – $800+ / mo $5k – $30k / yr $25 – $100 / mo
Recurring licence fees ✓ None ✗ Pay-as-you-go ✗ Annual subscription ✗ Monthly
REST API ✓ Full coverage ✓ ✓ ◉ Limited
Web Console / GUI ✓ ✓ ✓ ✓
VPC / Subnets ✓ ✓ ◉ Namespace-level ✗
Multi-tenant isolation ✓ Network-level isolation ✓ Account-level ◉ Namespace-level ✗
Network security (firewall) ✓ Security Rules, DDoS protection, TLS ✓ Security Groups ◉ NetworkPolicy ✗
Load Balancing ✓ ✓ ✓ ✗
Monitoring ✓ Built-in ◉ CloudWatch / extra cost ◉ Prometheus add-on ◉ Basic
Container orchestration ✓ ✓ ✓ ✗
Automatic scaling ✓ ✓ ✓ ◉ Limited
Auto-healing ✓ ✓ ✓ ✗
High Availability ✓ Multi-node ✓ Multi-AZ ✓ ✗
Zero-trust tunnels ✓ ◉ Add-on ✗ ✗
Run on your own hardware ✓ ✗ ◉ Extra cost ✗
No open inbound ports ✓ ✗ ✗ ✗
Persistent volumes ✓ ✓ ✓ ✗
Data sovereignty ✓ ◉ Region-locked ◉ Depends ✗ US only
Vendor lock-in ✓ None ✗ Proprietary ◉ Platform-specific ✗ Proprietary
Setup time ~15 min Hours–days Days–weeks Minutes
Total cost (3 years, 4 instances) $11.85 $14,400 – $28,800+ $15,000 – $90,000+ $900 – $3,600