Appearance & AI Models

Appearance Settings

Light and dark themes

Hypertask supports both light and dark themes, controlled from your profile or the Command Center.

• Dark theme â\x80\x94 reduces eye strain when working in low-light environments and is ideal for long coding sessions.
• Light theme â\x80\x94 provides high contrast for collaborative reviews and when sharing your screen.

You can toggle themes instantly from the settings menu. Preferences are saved per-user.

In AMOLED, floating AI chat has a thin outline and compact corners so it stays distinct from the black page, on desktop and phone. Other themes keep their existing appearance.

Keyboard shortcuts

From the Command Center, use Ctrl+K to open the global command palette. While focusing on a task detail, you can also use:

• Ctrl+O â\x80\x94 Open links from the latest reply in a new tab (Agent Chat and board comments).
• Alt+Left/Right â\x80\x94 Navigate through tasks or agents in increasing or decreasing order.

These shortcuts are available across boards, chat panels, and the Agents overview.

Tip

Some shortcuts are also active in the main command palette when a compatible view is loaded.

Text editor preferences

Comments and descriptions

When composing comments or editing task descriptions, you can:

• Dictate Ctrl+K, Ctrl+Space (on supported browsers): Start voice dictation from the text field.
• Save Ctrl+Enter: Send the comment; the task may stay or advance depending on context.
• Open links Ctrl+O: When a reply includes a clickable link, open it in a new tab.

The comment composer is pinned to the bottom of task detail panels, keeping it accessible while you read and compose.

AI models

Model selection

Hypertask supports multiple AI providers and models. During configuration, agents operate with provider-specific model names. Agent Chat respects the engine configured on an agent:

• Claude-based agents use Claude models (for example, claude-sonnet-5).
• Codex-based agents use Codex models (for example, gpt-5.6-luna).

The chat lane configuration must align with the agent's engine.

Note

You do not need OpenRouter-specific keys in Hypertask; the runtime handles exchange of credentials and model mapping. If you plan to use OpenRouter lanes beyond the default provider, include OPENROUTER_GLM_API_KEY for squad-based workflows.

Claude Haiku 5.5 is now Hypertask's default AI model for everyone, replacing Haiku 4.5. The new default is available across all plans, offering consistent model behavior for all users. When Haiku 5.5 is selected, it behaves equivalently to GPT 6 Luna in all supported surfaces including AI Task Writer, Agent Chat, and code-generation workflows.

Model quality now reflects real-world usage: Claude Haiku 5.5 cost per request is approximately 1.7× higher than GPT 6 Luna due to longer answer lengths, though the per-token price is identical. The model is fully operational on all supported environments, including Free and paid subscriptions.

Tip

Once the model is selected in your profile, the choice applies globally to any AI feature you use, keeping agent behavior and output style consistent across tasks, boards, and workspace contexts.

Quiet mode

Quiet mode controls how agent activity appears in run activities and thread comments. It applies to agents in active runs:

• Run activity â\x80\x94 by default, agents write detailed progress chatter to the run activity tab. Quiet mode suppresses most run activity logs.
• Thread comments â\x80\x94 quiet mode disables comment generation. Comments are only posted for specific actions:
  • Question
  • Decision
  • Handoff
  • Done

When running in quiet mode, agents remain fully engaged, but their directional footprint in the task history is streamlined. To enable quiet mode, use the quiet-run-activity token or configure it via board-level agent settings.

Tip

Quiet mode is useful for workflow histories where you want to review agent behavior via structured run logs, but keep thread conversations focused on human decisions and actions.

Icons, fonts, and visual tokens

Sidebar icons

Icon visual standards are managed in the Typography & Sidebar Visuals page. Controlled strokes and consistent spacing ensure icons are legible across light and dark modes.

Font scale

Typography uses a defined type scale for headlines, descriptions, and UI labels. Choose consistent weights and sizes from the standard set to maintain visual hierarchy.

Component reuse checklist

When adding UI controls to the app, define the existing component being reused or provide a specific reason. This checklist applies to all UI pull requests and is enforced via the canonical guide and required gate. Each new control must be mapped to an underlying component or documented with a no-reuse justification.

Note

Requests that reference or implement reusable components should include a mapping for each control added. Reviewers use the checklist to verify reusability claims and avoid unnecessary component proliferation.

---