Codecontainer

codecontainer

AI Development Environment for Coding Agents

Run Claude Code, Codex CLI, Gemini CLI, and OpenCode in isolated project containers.

`codecontainer` gives you a local AI development environment with safer defaults, persistent per-project containers, and built-in support for Docker, Podman, and Apple Container.

Use it when you want local AI coding, containerized development, and a practical sandbox for AI agents without building your own workflow from scratch.

The container image also includes browser automation tooling for agent-driven UI checks:

• Playwright with Chromium for deterministic headless E2E tests
• agent-browser for quick interactive smoke checks and debugging

Supports **Docker**, **Podman** (Linux), and **Apple Container** (macOS 26+) with automatic runtime detection.

Why This Exists

Running AI coding agents directly on your host machine gets risky quickly:

• they can modify the wrong files
• they can break local toolchains and dependencies
• they can access secrets and config you did not mean to expose

`codecontainer` solves this by creating a containerized development environment for each project, so AI agents can work locally with better isolation, repeatability, and safety.

Why Not Docker Alone?

Docker gives you containers, but it does not give you an opinionated local AI coding setup.

`codecontainer` adds the missing layer for AI-assisted development:

• per-project persistent containers
• agent-specific setup for Claude, Codex, Gemini, and OpenCode
• shared harness configs across projects
• optional yolo mode contained inside the sandbox
• browser testing tools available in the default image

How It Works

Host machine
  -> codecontainer
  -> Project container
  -> AI coding agent
  -> Your codebase

This makes `codecontainer` useful both as an AI agents sandbox and as a repeatable AI development environment for local coding workflows.

Use Cases

• Run Claude Code, Codex CLI, Gemini CLI, or OpenC