Project - Push-to-deploy hosting on dedicated machines

Overview

Hosting platforms have settled into two extremes: opinionated abstractions that hide the infrastructure entirely, or raw servers that demand a full platform team. Appentic sits in between. Every project runs on its own machine, in the geo you choose, on its own private network. You still just push to GitHub and it goes live.

I built the platform end-to-end. The control plane handles the GitHub App, project provisioning, deployment orchestration, and machine selection across six geos in Europe, the US and Asia. Services share the machine through a single memory slider, and CPU and disk follow the same share, so sizing is one decision rather than three.

The product surface covers the everyday tasks of running an app: web services, workers and cron jobs, managed PostgreSQL and Redis, S3-compatible storage, release commands and health checks, zero-downtime deploys with one-click rollback, automatic TLS and custom domains, live logs, a browser shell, and metrics with deploy markers. Billing is prepaid and charged by the second.

Appentic is also built for coding agents. Developers can connect Claude Code over MCP to price, provision, deploy and debug, with every billable action held until they confirm it. Athena, the built-in AI engineer, builds a Feather app from a plain description, tests it on the project's own machine and ships it. The platform itself runs on Feather (Flask + HTMX + Tailwind) and is built to be operable by a solo developer.

What I built