Repository navigation
AgentsInTheCloud
Coding agents that run on a laptop stop working when the laptop closes, and long tasks compete with everything else on the machine. AgentsInTheCloud moves that work to a server the user owns, so agents keep running while the user is away and can be checked from a phone.
The project site describes running coding agents remotely in isolated environments, with a fresh environment for each small task. It supports Claude Code and Pi, works with any model or subscription, and includes a mobile interface. The repository describes it as running cloud coding agents on the user's own server, with workflows optimized for evaluating agent work.
The site lists a Mac, a server at a provider such as Hetzner or DigitalOcean, or an old computer at home, connected through Tailscale. The README instructs the user to SSH into a Linux server dedicated to the project and run a one-line installer that pipes a remote script into bash. The script should be read before it is run on a server. The README does not describe the isolation mechanism or the architecture.
The repository was created on 2026-06-08 under the MIT license and had 160 stars and 4 open issues when checked, according to the GitHub API. It is maintained by Lucas Meijer, who states that contributions are not yet being accepted and that feedback is welcome.
- DockerSandboxes: managed isolation for agent execution.
- ParallelCodingAgentsTmux: running several agents at once on one machine.
- TailscaleSynology: Tailscale as the private access layer for self-hosted machines.
AgentsInTheCloud is placed in Assess at inner position. It addresses a clear problem, agent work that depends on a laptop staying open, and it is free and open source. The project is about four months old, and the README leaves isolation and architecture undocumented. There is no first-person production use, so the ring is Assess.
