Launching Portfork: host the apps your agents build

We are launching Portfork today.

If you work with AI coding agents for more than a few weeks, you run into a specific class of friction that is easy to feel and easy to ignore: your agents keep building apps, and every one of them runs on your laptop. A little dashboard here, a scheduled scraper there, an internal tool a teammate asked for. They work — right up until you close the lid. Then the URL is dead, the cron job never fires, and the thing your agent built is a demo, not a deployment.

Portfork fixes this. It is an on-demand host for the apps your agents build. Point it at an app, and Portfork keeps it running around the clock with a stable URL — on your own machine with Open Source Portfork, or on Portfork Cloud in an isolated microVM that scales to zero between runs.

The shape of the problem

The localhost trap is not a marketing slogan. It is a real cost paid in dead URLs, missed jobs, and infrastructure work that dwarfs the app itself.

Three concrete failure modes drove the design:

1. Apps die when you close the lid. Each app your agents build lives at localhost. Shut your laptop and it is gone — no URL to share, no uptime, no way for anyone else to reach it.
2. Deploying each one is a project. Containers, a host, networking, secrets, scaling — standing up real infrastructure for every one-shot app is more work than the agent spent building it.
3. Idle apps still cost you. Most agent-built apps run a few times a day. Leave a server up around the clock for each and you pay full price for capacity that mostly sits idle.

A traditional always-on host solves the first problem but doubles down on the third: you pay 24/7 for apps that run for ninety seconds a day. We rejected that design at the architecture stage.

How Portfork works

Portfork does three things:

• Places the app your agent built — packaging the artifact and putting it where it runs, locally on Open Source Portfork or on Portfork Cloud.
• Runs it 24/7. On Portfork Cloud the app lives in its own hardware-isolated microVM with a stable URL, staying up around the clock even with your computer off.
• Scales it to zero between runs. Apps wake on the first request and idle back down to nothing, so you pay for the node tier that fits your fleet — not for capacity you never use.

You drive all of it from one place. Portfork exposes a dashboard, a CLI, and an MCP server, so your agents can deploy, inspect, and manage their own apps without leaving the loop.

What ships today

The Open Source edition is shipping today, under AGPL-3.0:

• Lazy-start plus idle-reap local app hosting — apps wake on demand and reap when idle.
• Scheduled and agentic jobs you can run on a cron-style schedule.
• One control surface: dashboard, CLI, and MCP server.
• Unlimited local apps on your own machine.

The free edition runs entirely on your hardware. There is no markup on inference — hosted apps use your own model access, an API key or a claude setup-token subscription, directly.

What is coming

The Open Source edition is the foundation. Portfork Cloud builds on the same runtime and adds always-on hosting for people who want it:

• Cloud Starter, Plus, and Pro run your apps 24/7 in per-app isolated microVMs that scale to zero, priced by node tier.
• Each tier prices compute and hosting, not your model usage. The same artifact you run locally runs in the cloud — no rewrite, no migration.

The thread that ties it together is the runtime. The Cloud editions run the same app artifact the Open Source edition runs on disk. You choose where each app lives.

Why now

AI agents now build small apps faster than anyone can deploy them. The bottleneck moved: writing the app is the easy part, and getting it to run reliably somewhere your laptop is not, is the hard part. The right answer is a host that treats a ninety-second-a-day app as a first-class citizen — always reachable, never billed for idle.

That is what Portfork is for.

Get started

Install in one command. The free edition is enough to start.

curl -fsSL https://get.portfork.com/install.sh | sh

If you want always-on uptime, or a fleet of agent-built apps running 24/7, the pricing page covers what each tier adds. The same artifact runs locally and in the cloud — there is no migration.

Stop shipping demos.