Claude Code
Anthropic's CLI agent. Long-context, tool-using, file-editing. Run it inside the project Space and it sees your files, your CLAUDE.md, your tests.
Built in · Personal
TUI. Bring your CLI.
Run Claude Code, Codex, OpenCode, or any terminal AI tool — inside a Construct Space. Same project, same identity, same files. The CLI tools you already love, framed by the OS that gives them context.
What runs here
Bring the tool. Skip the setup.
Codex CLI
OpenAI's terminal coder. Same shell, same project, same identity layer underneath.
OpenCode
Open-source terminal agent. Plug it in, run it, treat it as another Space participant.
Anything that runs in a TTY
Vim, nvim, htop, k9s, Aider, your own tool. If it talks to a terminal, the TUI Space hosts it.
How it works
Real terminal. Real context.
Real PTY, not a fake
Full pseudo-terminal — colors, cursor control, mouse events, escape sequences. Apps behave like they would in iTerm.
Shared identity
Sign in once at the OS level. CLI tools that look for `.claude` / `.openai` / config files find them where they expect.
Project-scoped sessions
Open the TUI inside a project — the tool starts in the project root, with the project env, with the project files in reach.
Multi-pane
Run two CLI agents side by side. Compare outputs. Hand work between them.