RootTeam

RUNTIME FOR AI AGENTS

Give every agent its own machine.

Gentity spins up isolated AI agent instances on Fly Machines — each with its own URL, its own state, and its own runtime. Pick a template, click create, open the link.

Open dashboardView on GitHub
~/gentity
$ gentity create --template openclaw
→ Spawning Fly Machine in iad …
→ Mounting Volume agent-oc-x7k2 …
→ Reverse proxy ready
✓ https://oc-x7k2.gentity.app

WHAT IT IS

An agent, an isolated box, a URL.

Every Gentity instance is its own Firecracker microVM with a mounted Volume. Auth lives in front. State lives in back. You get a link.

Click → URL

Pick a template from the catalog. Click create. Get a dedicated subdomain that proxies to your Machine via authenticated edge.

Persistent state

Each instance has its own Fly Volume. Restart, stop, resume — agent memory, files, and config survive.

Cold start in seconds

Fly Machines boot Firecracker microVMs in under a second. Sleep when idle, wake on request. Pay for what runs.

HOW IT WORKS

Four steps. Nothing else.

No Docker Compose, no Kubernetes, no reverse proxy to babysit. Gentity composes Fly Machines, Volumes, and the edge proxy for you.

  1. 01

    Sign in with GitHub

    Allowlist-gated OAuth. Cookies are scoped to your subdomain tree so every agent inherits auth.

  2. 02

    Pick a template

    Choose an agent image and resource shape. Templates are typed: image, internal port, data path, env schema.

  3. 03

    Click create

    Gentity provisions a Fly Machine plus a Volume, persists the encrypted env, and registers the subdomain.

  4. 04

    Open the URL

    Hit <id>.gentity.app. The control plane authenticates the request and proxies through Fly's internal DNS to your Machine.

BUILT ON

Fly Machines

Firecracker microVMs, sub-second cold start

Fly Volumes

Persistent storage per agent

Next.js 16

Control plane + edge proxy

Auth.js

GitHub OAuth + allowlist

Prisma + Neon

Postgres for instance state

AES-256-GCM

Encrypted env per instance

Why this exists

Agents are not chat. Agents are processes. They need a home, a runtime, and an address.

Today every team building with AI ends up rewriting the same plumbing — a Dockerfile, a reverse proxy, an auth layer, a volume mount, a secret store. Then it breaks the moment two engineers want to run two agents.

Gentity is the missing layer between "I have an agent" and "anyone on my team can open it." Spawn an instance. Get a URL. Restart safely. Hand the link to someone. That's it.

The team behind Gentity is RootTeam — a Tokyo-born studio that's been operating SaaS for a decade and is now building AI-native infrastructure from Palo Alto.