Skip to main content
Smithers has two terminal UI surfaces: init’s one interactive question (skip with --yes / --non-interactive), and --interactive mode on up / workflow run, which opens a run launcher then the full-screen monitor. The standalone bunx smthrs tui command was removed in 0.20.2.

​
Interactive setup

Run init in a human terminal: it lists detected coding agents, asks which you prefer, installs the workflow pack with defaults plus its plugin (or skill if none exists), then opens a hijacked tutorial session hosted by that agent:
Pass --agent claude (or codex, pi, …) to skip the question and --no-tutorial to skip the guided session; for CI, scripts, or agents, use the non-interactive form:

​
Interactive run launcher

Use up --interactive to have Smithers pick a workflow and collect inputs. From an interactive TTY, omitting the workflow argument opens the same launcher:
Terminal screenshot of bunx smthrs up --interactive showing the workflow picker

Captured from tmux: up --interactive opens a searchable workflow picker before it starts a run.

Use workflow run --interactive when you prefer workflow IDs over file paths:
Passing a workflow ID skips the picker and goes straight to its input prompts:
Terminal screenshot of the interactive workflow input prompt

Passing an installed workflow id preselects it and prompts for its input schema.

​
Full-screen monitor

After the launcher starts the run, Smithers detaches the workflow and hands your terminal to the full-screen monitor (@smthrs/tui), which connects to the workspace Gateway (starting one if needed) and streams the run tree, output, logs, diffs, timeline, and hijack state.
Terminal screenshot of the full-screen Smithers TUI monitor showing a finished run

After a run finishes, the monitor still shows its run tree and selected node details.

Useful keys: If the TUI package can’t be resolved, the command fails loudly with TUI_MONITOR_UNAVAILABLE while the detached run keeps running; watch it with bunx smthrs ps, bunx smthrs inspect RUN_ID, or the run log file.

​
Command map

Agents should suggest interactive commands to humans but not run them through an automation harness; when executing a workflow itself, use -d, --format json, ps, logs, inspect, and events instead.