Meet Dieter
Your coding agents. Your machines. One native workspace.
Dieter brings coding agents running on your own machines into one workspace. Start a task on your Mac, let a Linux workstation do the work, and check the result from your phone. Every card and standalone chat keeps its own durable conversation.
Dieter is open source under the MIT license. The name is pronounced DEE-ter.
Start here
| You want to… | Go to |
|---|---|
| See what the apps can do | Product tour |
| Install a daemon and a client | Installation |
| Run your first task | Quick start |
| Organize projects, boards, and checkouts | Projects & tasks |
| Review files, changes, and running commands | Conversation workspace |
| Connect more machines | Machines & routes |
| Automate Dieter | CLI guide |
| Operate your own gateway | Self-hosting |
| Build or contribute | Development |
The product in a minute
Agents run on daemon hosts. Install dieter on Apple Silicon macOS or Linux
amd64/arm64. It starts agents beside your Git checkout using their normal local
configuration. Codex, Claude Code, Pi, Oh My Pi, and DeepSeek Harness are supported.
Native apps are your workspace. macOS, Android, and iOS provide boards, chats, files, terminals, schedules, and machine tools; Android and iOS share one app. The iPhone and iPad app is in beta; see the platform guide for its current distribution.
The gateway connects your devices. It authenticates accounts and enrolled machines, advertises routes, and relays end-to-end encrypted connections when needed, without being able to read them. Your devices trust each other only after you pair them. The gateway stores control metadata and normalized quota snapshots, not your repositories or conversations. There is no browser-based Dieter application; this website is documentation.
What stays where
Shared project metadata and portable settings replicate between your account’s daemons. Checkouts, conversations, schedules, files, and execution stay with their owning machine. Closing a client does not cancel agent work.
Agents have the permissions of the daemon user. Your chosen model provider may receive prompts, code, and tool output according to its own configuration. “Local execution” describes where the agent and tools run; it does not imply that model inference is local. Read the security model .