gptme-tui

gptme TUI — interactive terminal UI for gptme.

Complementary to the plain gptme CLI: supports queueing prompts while the agent works, collapsible tool output, and a live status bar.

PROMPTS are submitted on start, like with gptme; chain several with ‘-’ (e.g. gptme-tui “write a script” - “now test it”).

Usage

gptme-tui [OPTIONS] [PROMPTS]...

Options

-n, --name <name>

Conversation name to open or create.

-r, --resume <NAME>

Resume a conversation: NAME if given, else pick one from a list.

-m, --model <model>

Model to use (e.g. anthropic/claude-sonnet-4-6).

-w, --workspace <workspace>

Workspace directory (default: current directory).

-t, --tools <tool_allowlist>

Comma-separated list of tools to allow.

--tool-format <tool_format>

Options:

markdown | xml | tool

--no-confirm

Skip tool confirmation prompts.

--inline

Experimental: render inline without the alternate screen, printing messages into the terminal’s native scrollback (tmux/terminal scrolling works normally, like Claude Code).

--experimental-jelly-errors

Animate TUI error messages and show recovery hints.

-v, --verbose

Enable verbose logging.

Arguments

PROMPTS

Optional argument(s)