We’re building Ever Works in public. Some things may be incomplete, missing, or broken while we continue improving the platform.We’re building Ever Works in public — expect a few rough edges.
Feature · Local-first
Ever Works does not have to live in someone else’s cloud. The Desktop app runs the whole platform on your own machine, your laptops and servers can take Agent work as Fleet nodes without opening a single port, and everything the platform produces is committed to repositories you own. Ever Works does not watch people; this page is about where the platform runs and where your data lives.
Pick one, or mix them. Where the platform runs and where its work executes are separate choices.
The API, the dashboard, the job runtime and the database run as one app on your machine. Embedded SQLite needs no other services, and the tray keeps the stack running with the window closed. It is early access: take an installer from a CI build or build it from source, because there is no public download page yet.
Enroll a laptop, workstation or server with a one-time token and it can take Agent runs. Enrollment is outbound-only, so the node calls the platform and never the other way round. Nothing on the machine has to be reachable from the internet.
Run the full platform on your own servers from the published Docker images with the Compose files, or deploy it to Kubernetes from the manifests in the repository. The code is open source under AGPLv3.
Point the AI provider at a model running locally: the Ollama and LM Studio plugins talk to a local server, and neither needs an API key by default.
ever-works-cli talks to the same REST API as the dashboard, including your own instance via --api-url. Credentials go into a file only you can read, and the CLI refuses to send a token over plain http:// unless the host is loopback.
The local stack itself needs no cloud backend. The providers you plug in still sit where they sit, though: pushing to your Git host and deploying to your target go over the network exactly as they do in the cloud, and a hosted AI model needs its API. What the platform does document is how it behaves when one of your machines drops off the network.
Running on your own machines is the operational half of the ownership promise: the work happens on hardware you control, next to repositories you control. Start with the Desktop app on one machine, add nodes as the workload grows, and point at your own cluster when it makes sense.