🔒 Run AI coding agents inside isolated, project-specific development environments.
AIDC, short for AI Dev Container, is a one-command wrapper for running AI coding agents inside hardened Linux containers. It supports Claude Code, Codex, OpenCode, Grok, and Cursor Agent while limiting what each agent can access on the host system.
Instead of giving an agent access to your complete home directory, shell history, SSH keys, and unrelated projects, AIDC mounts the selected project at /workspace and keeps agent state inside dedicated Docker volumes.
With AIDC, you can:
✅ Create a dedicated development container for each repository
✅ Run multiple AI coding agents through a consistent command-line interface
✅ Keep host SSH keys, shell history, and unrelated files outside the agent environment
✅ Persist agent configuration and session state in per-project Docker volumes
✅ Detect project languages and install relevant toolchains automatically
✅ Run built-in security tools such as Semgrep, Gitleaks, TruffleHog, Bandit, gosec, cargo-audit, and ShellCheck
✅ Add software supply chain checks and package-management guardrails
✅ Enable a default-deny egress firewall when tighter network control is required
✅ Use dedicated per-project virtual machines through Lima on macOS, with experimental Firecracker support on Linux
Why it stands out:
🧱 Project-specific isolation instead of unrestricted host execution
🔐 No full home-directory mount
🛡️ Security scanners and guardrails are included in the development environment
📦 Agent state remains persistent without exposing unrelated host data
🧰 Supports multiple AI coding agents and common development toolchains
AIDC is currently a pre-1.0, rolling-release project. It is macOS-first and requires Docker Desktop, OrbStack, or Colima. Interfaces and behavior may change as the project evolves.