Settings Reference

Accessing Settings

Open Settings at any time by pressing Command + , or clicking the gear icon in the top bar. Settings are organized into tabs, each covering a specific area of the app.

General

The General tab controls core interaction preferences.

• Font size — Adjusts the font size for the chat panel and message text. Choose a size that is comfortable for extended reading and writing.
• Send shortcut — Select how messages are submitted to Claude Code. Options are Command + Enter (default) or Enter. When set to Enter, use Shift + Enter to insert new lines in your message.
• Default session mode — Sets the mode that new sessions start in. Options are Normal (approve each action), Auto-Accept (run without pausing), Plan (reasoning only, no code changes), Auto (CLI auto-routes per tool), Don't Ask (accept everything without prompts), or Bypass (skip all permission checks — dangerous). You can always change the mode within a session with Command + . or Shift + Tab. See Session Modes Explained for details.
• Default thinking effort — Persisted as defaultThinkingEffort. There is no stand-alone control on this tab; the value is set from the Effort selector in the Input Area (see Chat Panel) and is passed to every new Claude CLI session via --effort. Valid levels come from the running model's capability manifest (e.g. low / medium / high, plus xhigh on the Default model). Set to null to inherit whatever Claude Code's own config specifies.

AI Providers

The AI Providers tab manages API keys for the multi-AI assistant feature. These keys are used by the assistant sidebar, not by Claude Code itself (which uses your Anthropic subscription directly).

• OpenAI API key — Enables GPT-based assistants in the assistant panel.
• Google Gemini API key — Enables Gemini-based assistants.
• Anthropic API key — Enables additional Claude-based assistants separate from your Claude Code session.

API keys are stored locally in your application data directory and are never sent to CodeMantis servers. For a detailed walkthrough of setting up and testing API keys, see Setting Up AI API Keys.

Terminal

The Terminal tab configures the built-in terminal emulator.

• Default shell — The shell that opens when you launch a terminal tab. Defaults to your system shell (typically /bin/zsh on macOS).
• Terminal font size — Independent font size control for the terminal, separate from the chat font size.

Changelog

The Changelog tab controls automatic changelog generation for your projects.

• Provider selection — Choose which AI provider generates changelog entries (uses your configured API keys from the AI Providers tab).
• Auto-generate toggle — When enabled, CodeMantis automatically generates a changelog entry after each session that modifies files. Disable this if you prefer to manage changelogs manually.

Appearance

The Appearance tab controls the visual theme.

• Theme selection — Choose from six built-in themes: Midnight, Ocean, and Ember (dark), or Dawn, Sand (default), and Arctic (light). The default theme is Sand on a fresh install. Changes apply immediately. See Themes & Appearance for details on each theme.

Shortcuts

The Shortcuts tab provides a quick reference for all keyboard shortcuts available in CodeMantis. This tab is read-only — it displays the current shortcut bindings for easy lookup. See Keyboard Shortcuts for the full reference.

Session Logs

The Session Logs tab controls conversation persistence and retention.

• Save session conversations — When enabled (default), CodeMantis saves full conversation history to a local SQLite database. This includes messages, tool operations, and approval decisions.
• Retention period — How long saved sessions are kept before automatic cleanup. Options: 7 days, 14 days, 30 days (default), 90 days, or Forever. Cleanup runs automatically when CodeMantis launches.

For a detailed walkthrough, see Session Persistence & Chat Logs.

Preview

The Preview tab configures the built-in preview browser.

• Preview window dimensions — Sets the default width and height for the preview window (default 1024x768).
• Auto-start dev server — When enabled, CodeMantis automatically starts a dev server when a project is opened. Useful for web projects where you always want a running server.
• Custom dev command — Override the detected dev command with a custom one (e.g. npm run dev -- --port 4000).
• Auto-open console on errors — When enabled, the console panel in the preview browser opens automatically when JavaScript errors are detected.

SpecWriter

The SpecWriter tab configures the behavior of the SpecWriter feature.

• Default mode — Sets whether SpecWriter opens in Guided mode (step-by-step questionnaire) or Freeform mode (open text editor) by default. You can switch between modes within SpecWriter at any time.
• Auto-save preference — When enabled, SpecWriter automatically saves your specification as you edit. When disabled, you must save manually. Auto-save is recommended to prevent accidental data loss during long specification sessions.
• Max output tokens — Controls the maximum length of generated specifications (range: 1,024–200,000 tokens). Higher values allow for more detailed specs but use more API credits.

Super-Bro

The Super-Bro tab configures the proactive guidance assistant.

• Enable Super-Bro — Master toggle. On by default. Super-Bro stays quiet unless it has something actionable to say, so it ships enabled.
• Provider — AI provider for Super-Bro evaluations. Options are "Auto (cheapest available)", OpenRouter, Google Gemini, OpenAI, or Anthropic. Only providers with configured API keys appear.
• Model — Specific model. "Auto" picks the cheapest available model automatically.

For details, see Super-Bro: Proactive AI Guidance.

Self-Drive

The Self-Drive tab configures the autonomous implementation orchestrator that automates Implementation Guide sessions.

• Provider — The AI provider for the orchestrator (Anthropic, Google Gemini, OpenAI, or OpenRouter). Only providers with configured API keys are listed.
• Model — The specific model for the orchestrator. A fast, inexpensive model (Haiku-class) is recommended — the orchestrator only reads output and makes decisions.
• Max fix attempts per session — How many times Self-Drive retries a failing session before pausing (range: 1-10, default: 3).
• Run build check after each session — When enabled (default), Self-Drive runs the project build after each session to catch compile errors.
• Run test suite after each session — When enabled (default), Self-Drive runs the test suite after each session to catch regressions.
• Auto-commit between sessions — When enabled, Self-Drive creates a git commit after each successfully completed session. Off by default.
• Recheck loop (instead of pausing on format gaps) — When enabled (default), the orchestrator can ask Claude Code to re-state evidence for specific verify items before pausing — up to 2 rounds per session, with at most 1 recheck per item. Disable if you want to review every borderline verification pause manually.

The default Self-Drive Provider is Anthropic and the default Model is Claude Haiku 4.5 (claude-haiku-4-5). Both are auto-selected from the providers with a configured API key.

For a detailed walkthrough, see Self-Drive: Autonomous Implementation.

Assistant

The Assistant tab configures multi-AI assistant behavior.

• Shortcut buttons — Configure reusable prompt shortcuts that appear as quick-access buttons in assistant tabs.
• Default model — Set the default AI model used when creating a new assistant tab.

Quick Commands

The Quick Commands tab lets you customize the terminal's quick command buttons. Each command has:

• Label — The text shown on the button (e.g. "Build", "Test")
• Command — The shell command to execute (e.g. pnpm build, cargo test)

Add, edit, or remove commands to match your project's workflow. See Integrated Terminals for details.

Shortcuts & API Logs

• Shortcuts — A read-only reference of all keyboard shortcuts. See Keyboard Shortcuts for the full list.
• API Logs — View and filter API request/response logs for debugging. Each log entry shows the request method, endpoint, status code, and timing. Click an entry to inspect the full request and response payloads.