Pinned Prompts

Pinned prompts are named shortcuts that appear as clickable pills above the chat composer. Use them to send recurring instructions, slash commands, or multi-line prompts in a single click — without retyping or scrolling through history.

The Prompt Bar

When at least one pinned prompt is configured, a row of pills appears above the chat input. Each pill shows the prompt’s display name. Clicking a pill either sends the prompt immediately or inserts it into the composer, depending on whether Auto-send when clicked is enabled.

Auto-send vs Insert

Each pinned prompt has an Auto-send when clicked toggle (referred to as Auto-send below for brevity):

Mode	Behavior
Auto-send on	Clicking the pill sends the prompt immediately, as if you had typed it and pressed Enter
Auto-send off	Clicking the pill inserts the prompt text into the composer so you can review or edit before sending

Use auto-send for prompts you always want to run as-is. Leave it off for templates you customize each time.

Toolbar Overrides

Each pinned prompt can also carry an opinion about the four chat-toolbar toggles:

• Plan mode
• Fast mode
• Thinking
• Claude in Chrome

Each override is tri-state: Inherit, On, or Off.

The Plan mode override follows the active runtime: Claude-compatible sessions start with Claude Code’s plan approval flow, while Codex Native sessions start with Codex’s plan collaboration mode.

Setting	Behavior on click
Inherit (default)	The toolbar toggle is left at whatever you’ve selected for the session
On	The toolbar toggle is flipped on before the prompt runs
Off	The toolbar toggle is flipped off before the prompt runs

Overrides are applied to the live toolbar — the change is sticky, not per-turn. After clicking the pinned prompt, the toolbar shows the new state and follow-up turns inherit it until you flip the toggle back yourself.

A typical use: an “Incorporate feedback” prompt that must run outside plan mode can pin Plan mode: Off, so clicking the pill always disables plan mode before sending — even if you forgot the toolbar was still on plan.

Overrides apply whether Auto-send is on or off. With auto-send off you’ll see the toolbar update before you edit the inserted prompt.

Launch Options

Beyond the toolbar toggles, a pinned prompt can decide where it runs and on which model. Both live under Launch options in the prompt editor and are independent of each other.

Open in a new tab

With Open in a new tab enabled, clicking the pill:

1. Creates a fresh chat session in the current workspace
2. Names that tab after the prompt’s display name
3. Applies the prompt’s toolbar overrides and model to it
4. Switches to it and runs the prompt there

The tab you clicked from is left untouched — its conversation, toolbar state, and any in-progress turn all continue undisturbed.

Because the tab is explicitly named after the prompt, Claudette’s automatic session naming is skipped for it. You can still rename it by double-clicking the tab.

With Auto-send off, the new tab opens with the prompt text already in its composer instead of sending it.

Model

The Model dropdown selects a model before the prompt runs. Leave it on Inherit (don’t change) — the default — for the prompt to use whatever model the target session already has.

Combination	What happens
Model + new tab	The fresh tab starts on that model. Nothing to migrate, so switching backends here is free
Model, no new tab	The active session switches models before the prompt runs — exactly like the /model slash command, including its context handling

Switching model in the active session follows the same rules as the toolbar picker: a swap within the same runtime (say Opus to Sonnet) keeps the conversation, while a swap across runtimes (say Claude Code to Codex) carries the prior turns over as a summary prelude.

If a pinned prompt’s model becomes unavailable — you disabled its backend, or the provider stopped offering it — the prompt still runs, on the session’s own model. The editor keeps your choice and flags it as unavailable so re-enabling the backend restores it.

The review-in-Codex workflow

A common setup: a global Review prompt with the body /code-review, Auto-send on, Open in a new tab on, and Model set to a Codex model. One click on the pill while you’re mid-implementation opens a separate, named review tab on a second opinion — without disturbing the session doing the work.

Using ”/” in Prompt Bodies

When creating or editing a pinned prompt, the prompt body field supports the same / slash command picker as the main chat composer. Type / to open the picker and select a command to insert it.

This lets you build prompts that combine slash commands with custom context — for example, a “Review changes” prompt that runs /review followed by project-specific instructions.

Global Prompts

Global pinned prompts appear in every repository. Manage them at Settings → Pinned Prompts.

To create a global prompt:

1. Open Settings → Pinned Prompts
2. Click Add prompt
3. Enter a Display name (shown on the pill) and the Prompt body
4. Toggle Auto-send on or off
5. Under Toolbar overrides, set each toggle — Plan mode, Fast mode, Thinking, Claude in Chrome — to Inherit, On, or Off (see Toolbar Overrides above; leave them on Inherit if the prompt doesn’t care)
6. Under Launch options, optionally enable Open in a new tab and pick a Model (see Launch Options above)
7. Click Save

Use the ↑ and ↓ buttons to reorder prompts within the list. Order is preserved in the pill bar.

In the prompt list, a prompt that forces any toggle shows small chips (e.g. Plan mode: Off) next to its name, alongside the Auto-send indicator. Launch options get their own chips — New tab and Model: <id>.

Repo-Specific Prompts

Each repository can have its own pinned prompts that appear only when working in that repo. Manage them at Settings → [Repository Name] → Pinned Prompts.

Repo prompts appear before global prompts in the pill bar. See Per-Repo Settings for other repository-level configuration.

Inheritance and Overrides

Repositories inherit all global pinned prompts automatically. If a repo defines a prompt with the same display name as a global prompt, the repo’s version takes precedence and the global one is hidden.

In repo settings, inherited global prompts are listed below the repo-specific prompts. Any global prompt that is being overridden by a repo prompt of the same display name shows an Overridden badge.

This lets you define a common baseline globally while tailoring specific prompts per project — for example, a global “Run tests” prompt that a repo overrides with its specific test command.