AI is making software abundant.
We need a better way to manage it.
AI coding tools have made it trivial to build tiny, highly specific applications. The problem is what happens afterward.
Habitat is the private application layer for the software organizations build themselves.
Currently exploring this with founders and AI-heavy engineering teams.
Your software
Interactive concept. Hover an app, filter by department, click for details.
GitHub manages your code.Habitat manages the software you build.
We are going to have more software than we know what to do with.
- Build software→
- Deploy it→
- Maintain it
- Build software→
- Build another→
- Build another→
- Build another…
AI made software cheap. Now we're going to have too much of it. Forgotten projects, random folders, repositories nobody remembers, deployments nobody owns, useful tools nobody knows exist.
The cost of creating software is falling below the cost of organizing it.
This is what the software looks like today.
Desktops, downloads, forgotten repos, preview deployments, a cron job on someone's laptop. Real software, doing real work, with no home.
Chaos → Application layer
Same software. Now it has a name, an owner, a department, a status and a private URL.
Most infrastructure assumes software should be public.
A developer builds a supplier price analyzer for three people in procurement. The URL should not be the primary primitive. The organization should be.
- 01BuildClaude writes it in an afternoon.
- 02DeployPush to a hosting platform.
- 03Public URLexposedNow it's on the internet.
- 04Add authenticationLater. Maybe.
- 05Configure networkingAllowlists, tunnels, env vars.
- 06Manage accessWho has the link? Nobody knows.
- 01BuildClaude writes it in an afternoon.
- 02DiscoverHabitat notices it exists.
- 03Private by defaultinternal onlyReachable only inside your network.
- 04SSOYour identity provider, automatically.
- 05Share with the right peopleProcurement. Not the internet.
Public is a choice.
Private is the default.
Private networking doesn't magically solve security. It dramatically reduces unnecessary public exposure and attack surface. Public access becomes an explicit decision instead of an accident.
AI creates. Habitat does everything after.
No ports, no URLs, no deployment configs, no auth middleware. The developer builds. The organization gets an application.
- 01Build
Claude, Cursor or Codex creates the application.
$ claude> Build me a tool that compares supplier prices.✓ Created supplier-analyzer - 02Discover
Habitat automatically finds it.
scanning dev-macbook-rishinew applicationsupplier-analyzer~/code/supplier-analyzer · Node.js · port 3000 - 03Run
Habitat figures out how to run it.
runtimenode 22startnpm run startdepsinstalledhealth200 / - 04Share
Give it a stable private URL.
private networkapps.acme.internal/suppliersshared with Procurement · 3 people - 05Govern
SSO, permissions, ownership, health and lifecycle.
SSO · OktaOwner · RishiUsers · 8HealthyDeps · current
From “I built a tool” to “my company has an application.”
The app runs where it was built, or wherever Habitat runs it. No public endpoint was created.
No ports, no DNS, no auth middleware. The developer never left the terminal.
A company application, with an owner, users, a status and a private URL.
Every internal application becomes a first-class company application.
Compare vendor pricing across quotes
- Owner
- Rishi
- Users
- 8
- Health
- Healthy
- Access
- Procurement Private·SSO
- Repository
- acme/supplier-analyzer
- Runtime
- Node.js 22
- Last updated
- 2h ago
- Built with
- Claude Code
Enterprise deployments can automatically expose applications through the organization's identity provider, putting SSO and access control in front of software without requiring every developer to build it themselves.
Software used to be scarce.
AI is making it abundant.
The bottleneck is moving.
When software becomes cheap enough to create for every niche workflow, organizations need a new layer for discovering, running, sharing, securing and maintaining what they've built.
One home for everything your organization builds.
We're not pretending the product does all of this today. This is the shape of the opportunity if AI-generated internal software becomes as common as we think it will.
- Discovernow
Find applications across machines, repos and infrastructure.
- Runnow
Understand how it runs and run it. No manual ports or configs.
- Sharenow
Stable private URLs over the organization's network or VPN.
- Securenext
Private by default. SSO through your identity provider.
- Monitornext
Know what's healthy. Let users report what isn't.
- Maintainlater
Detect broken deps, expired credentials, drift. Let agents fix them.
Is this actually a problem?
We're testing this thesis with founders, CTOs and teams building heavily with AI.
We want the honest version, especially if you think this is bullshit.
- 01Do you already have a pile of small internal software?
- 02How do you currently run and share it?
- 03Do you feel like you have software you built but can no longer find?
- 04Would private-by-default deployment + SSO be useful?
- 05Which part of this feels most painful?
- 06Which part feels like a solution looking for a problem?