Customize Prompt Output Settings

Control what appears in your generated prompt — screenshots, ARIA trees, DOM diffs, agent target format, and maximum recording length.

Output settings let you control what Claude Scope includes in the prompt it generates from a recording. You can set global defaults that apply to every new recording, and override them per session on the Output page.

Accessing output settings

Go to your workspace and click Settings in the left sidebar, then scroll to the Output Preferences section.

Available settings

Setting	Default	Description
Default agent target	Claude Code	Determines the formatting and framing of the generated prompt. Options: Claude Code, Codex, Cursor, Raw.
Include screenshots	On	Embeds extracted frame thumbnails inline in the generated prompt.
Inline ARIA tree	On	Appends the full accessibility snapshot to the prompt, giving the agent a structural view of the UI.
Include raw DOM diff	Off	Adds an element-level diff between consecutive frames, showing exactly which DOM nodes changed.
Max recording length	5 minutes	Caps how long a recording session can run. Options: 5, 10, 15, or 30 minutes.

Setting descriptions

Default agent target

The agent target controls how Claude Scope frames and structures the generated prompt:

• Claude Code — Formats output for use with Anthropic’s Claude Code CLI agent.
• Codex — Formats output for use with OpenAI Codex.
• Cursor — Formats output for pasting into Cursor’s system prompt or composer.
• Raw — Outputs the prompt with no agent-specific framing, suitable for any model or custom tooling.

Include screenshots

When enabled, Claude Scope embeds a thumbnail of each extracted frame directly in the prompt body. This gives the agent visual context for each UI state captured during the recording.

Disable this setting if you want a text-only prompt, or if your agent does not support image input.

Inline ARIA tree

When enabled, the full ARIA accessibility snapshot captured by the Playwright analysis lane is included in the prompt. This gives the agent a structured, semantic view of the UI — component roles, labels, and hierarchy — without relying solely on visual inspection.

Keeping the ARIA tree enabled alongside screenshots gives agents the richest context. Disable it only if you are working with models that have tight token limits.

Include raw DOM diff

When enabled, Claude Scope appends an element-level diff between each pair of consecutive frames. The diff shows which DOM nodes were added, removed, or modified between UI states.

This setting is off by default because the diff output can be verbose. Enable it when you need the agent to reason about fine-grained state changes, such as validating form behavior or tracking dynamic content updates.

Max recording length

Sets the maximum duration for a new recording session. When the limit is reached, Claude Scope stops the recording automatically. Choose a longer limit for complex workflows, or keep it short to stay focused.

The demo environment uses a 30-second maximum regardless of this setting. The full range of options (5, 10, 15, 30 minutes) applies to standard accounts.

Per-session overrides

The settings above are saved to your account and applied as defaults for every new recording. On the Output page for any individual session, you can override each setting before copying or downloading the generated prompt. Overrides apply only to that session and do not change your saved defaults.