Coding agents in containers

qcode

Set up a coding agent once, sign it in, and open it in any workspace in seconds. Agents and shells sit side by side as tabs, and everything an agent runs stays inside the project's container.

betagithub ↗

Rust
Podman
Docker
Claude Code
Codex

qcode workspace with agent and shell tabs next to the project file tree

notes

qcode started as a desktop IDE for multi-agent work and ended as a terminal application, which turned out to be the better shape: it runs anywhere I have a shell, including over SSH.

Each project gets its own container. Agents, shells and files open as tabs, the home screen restores the projects and tabs you had open, and when the last qcode window closes the containers it started are stopped too. The design goal is simple: an agent that goes wrong should damage a disposable filesystem, never the machine it runs on.

facts

crate: quvyta-code
agents: Claude Code · Codex · Gemini CLI · opencode
isolation: Podman / Docker containers
built on: quvyta-framework

Screenshots

qcode home screen restoring recent projects — home: projects and tabs come back after a restart

qcode file manager inside a project container — files: managed inside the project container

qcode listing the coding agents it can run — agents: set up once, open anywhere