Why Codius?

Codius is a self-hostable platform for running and orchestrating coding agents. It runs the agent CLIs you already use, on the hardware you already have, and gives you the App, Codius CLI, and API to drive them from anywhere.

Architecture

• Daemon-client architecture. The daemon manages agents; the mobile, desktop, and web Apps plus Codius CLI connect locally or over a relay. Remote access isn't an add-on.
• macOS, Windows, and Linux are all primary targets. None of them are a port or an afterthought.
• Mobile, desktop, and web are separate native clients. The mobile app is built in React Native, not a webview.

Providers

• Bring your own. Use your Claude subscription, your OpenAI account, your own API keys, a self-hosted endpoint. Codius doesn't proxy model calls.
• Local voice stack. Speech-to-text and text-to-speech run on-device by default. OpenAI providers are configurable if you want cloud quality.
• Open source. No telemetry on your code.

Where agents run

• Your laptop, a homelab, a company server. Same daemon, same client surface.
• Any directory, git or not. Launch agents, merge locally, review the diff in the app.
• GitHub PRs, checks, and reviews surface in the app when you want them. Not required.

Parallel work

• Splits and panes. Agents, terminals, and browsers side by side in one workspace.
• Per-worktree services. Each worktree gets allocated ports for dev servers and databases, reachable through proxy URLs like web.fix-auth.my-app.localhost so they don't collide.
• Multiple agents on the same repo via worktrees.

Automation

• Codius CLI exposes the same host-management surface for scripts and automation.
• Codius tools. Agents can drive Codius themselves: create isolated workspaces, spawn subagents, open terminals, and send prompts.

What it isn't

Not a hosted agent, not an IDE, not a model provider. Codius runs the CLIs you already use and stays out of the way.