Skip to main content

Documentation Index

Fetch the complete documentation index at: https://www.rumus.ai/docs/llms.txt

Use this file to discover all available pages before exploring further.

Shell integration is what lets Rumus know where one command ends and the next begins. With it on, you get colored decorations for command success or failure, accurate working-directory tracking, and tighter AI agent behavior. Without it, terminals still work — you just lose those niceties.

​
What you get when it’s on

Command decorations

A small circle in the gutter next to each command line — purple for success, red vertical line for non-zero exit.

Working directory

Rumus knows your current directory at any moment, which improves AI suggestions and tab tooltips.

Cleaner prompt boundaries

Search and selection respect command boundaries — you can scroll back and reach “the previous prompt” precisely.

Better AI context

The agent uses prompt and exit-code markers to understand what just ran and how it ended.

​
How it works

Shell integration is built on the standard OSC 133 escape sequences — the same convention VS Code, WezTerm, and friends use:
MarkerMeaning
OSC 133;APrompt start
OSC 133;BCommand start (after the prompt)
OSC 133;CCommand execution begins
OSC 133;D;exitcodeCommand finished, with exit code
When Rumus sees these markers, it draws decorations and updates the working directory.

​
Turning it on

Shell integration is configured per profile in Settings → Profiles. Each profile has a toggle — labeled something like “Enable script injection for command decorations and working directory detection” — that injects the necessary markers into the shell when a tab using that profile starts. Built-in profiles (the bash/zsh/PowerShell entries Rumus auto-detects) ship with this on. If you’ve created a custom profile that wraps a non-standard shell, you have two options:
  • Leave the toggle on and let Rumus inject standard markers.
  • Turn the toggle off and add OSC 133 markers manually in your shell’s startup config.

​
Supported shells

The auto-injection covers the major shells:
  • bash and zsh — markers added to your prompt via PROMPT_COMMAND / precmd.
  • PowerShell — markers added to your prompt function.
  • fish — markers added via fish_prompt event handlers.
  • cmd — limited; some markers are not feasible due to the shell’s prompt model.
WSL distributions inherit whichever shell you use inside them.

​
When it doesn’t work

A few things can break shell integration:
  • A custom prompt that overwrites the markers. If you have your own PROMPT_COMMAND or precmd that doesn’t preserve Rumus’s hooks, decorations stop appearing. Make sure your prompt logic is appended rather than replacing the existing one.
  • Connecting to an old shell over SSH. Older bash versions or minimal dash shells may not support all the prompt hooks needed.
  • Programs that take over the screen. Full-screen TUIs (vim, htop, k9s) intentionally suppress prompts; decorations resume after the program exits.
If you don’t see decorations and you expected to:
  1. Open a new tab using the affected profile.
  2. Run a quick command (ls).
  3. Confirm the toggle is on for that profile.
  4. Check that your shell’s RC file isn’t overwriting the integration hooks.

​
Privacy

Shell integration runs entirely on your device. The markers it injects are local-only — they’re inserted by your shell, parsed by your terminal, and never sent anywhere.
Shell-integration not behaving the way you expected? Ask in the Rumus community — the more shell variations we hear about, the better the integration gets.

​
Next steps

Profiles

Configure the per-profile toggle and add custom shells.

Search

Search across the prompt boundaries that shell integration creates.