# style and statusline

> tokenade style sets a short note that keeps your agent's replies lean; tokenade statusline shows tokens saved in the agent's status bar.

Source: https://tokenade.net/en/docs/style-statusline · Updated: 2026-10-07

`tokenade style` controls a one-line note Tokenade adds ahead of your prompt so the agent answers without greetings, recaps or narration. `tokenade statusline` controls the line in your agent's status bar that shows how many tokens Tokenade saved. Both are global settings.

## style

```text
tokenade style                       # show the current setting
tokenade style <auto|chat|coding|off>
```

| Value | Effect |
| --- | --- |
| `auto` | Default. Picks `chat` or `coding` per prompt from what you typed. With no clear signal, nothing is added |
| `chat` | Concise, human-readable answers: answer directly, no greetings, no recap of the request, no narration, no closing summary |
| `coding` | For multi-tool coding work: one ASCII status line per tool call (verb · target · outcome), identical findings grouped with `(×N)` |
| `off` | Never touch the agent's voice. Use it for benchmarking raw model behaviour |

```text
$ tokenade style
current style: auto (set with `tokenade style auto|chat|coding|off`)
  auto    (default) pick chat/coding per prompt automatically
  chat    concise human-readable answers
  coding  one dense ASCII status line per tool call
  off     never touch the agent's voice
```

Each note is about 50 tokens. It starts with `[tokenade:style=chat]` or `[tokenade:style=coding]`, so you can spot it in a transcript.

### Where it is stored and how it is sent

- The setting is global: `~/.tokenade/style.json` (override the path with `TOKENADE_STYLE_PATH`). A `.tokenade/style.json` left in a project by an older version is moved to the global file.
- On Claude Code, the UserPromptSubmit hook adds the note on the first turn, when the style changes, and after the conversation is compacted, not on every prompt.
- Gemini-family agents get it from their hooks; agents without a prompt hook get it from their rules file or from the [LLM proxy](https://tokenade.net/en/docs/llm-proxy), never from both.
- A change takes effect on the next prompt; nothing needs restarting.

## statusline

```text
tokenade statusline [--json|off|on|status]
```

`tokenade install` sets Claude Code's `statusLine` to run `tokenade statusline`, and wires the same ticker into Copilot CLI, Cursor CLI and Qwen Code. During a session it shows tokens saved in that session and all-time, with an estimated money value. A session that saved nothing shows nothing.

| Subcommand | Effect |
| --- | --- |
| `off` | Opt out. Removes the ticker and restores your previous status line command |
| `on` | Re-enable it |
| `status` | Show whether the ticker is on |
| `--json` | Print `saved_tokens`, `tokens_before`, `tokens_after`, `percent`, `usd`, `model` |

```text
$ tokenade statusline status
tokenade statusline ticker: ON
```

### Your own status line is kept

If you already had a custom `statusLine` command, install keeps it: your command runs first and prints its lines, and the Tokenade ticker goes on its own line below. `statusline off` puts your original command back. If you remove or replace the line by hand, Tokenade does not add it again.

Environment switches, for scripts and managed machines:

| Variable | Effect |
| --- | --- |
| `TOKENADE_NO_STATUSLINE=1` or `TOKENADE_STATUSLINE=0` | Ticker off |

The `off` choice is stored as a marker file in `~/.tokenade/` (`statusline-off`), so it survives upgrades.

## Related

- [gain, dashboard and receipt](https://tokenade.net/en/docs/gain-dashboard-receipt)
- [How hooks intercept tool calls](https://tokenade.net/en/docs/hooks)
- [The LLM proxy](https://tokenade.net/en/docs/llm-proxy)
- [Claude Code (CLI and desktop)](https://tokenade.net/en/docs/claude-code)
