claude-code-haha/docs/en/desktop/settings.md

title, nav_title, description, order

title	nav_title	description	order
Settings guide	Settings	Choose the settings needed for a first session, then find advanced options by task.	6

Settings guide

Click Settings at the bottom of the sidebar. For your first session, connect a model in Providers, check permissions and language in General, then follow Your first session. Open the other tabs when a task calls for them.

Providers

Model access. Sign in to Claude, ChatGPT, or Grok with an account (no API key required), or add any Anthropic- or OpenAI-compatible service with an API key.

You'll come here once during setup and rarely again. Full steps in Connecting a model.

General

General covers four areas: appearance and replies, how the agent works, network and notifications, and local data. Check the first two before your first session; adjust the others when you need them.

Before your first session: permissions and replies

Setting	What to choose	When it takes effect
Default session permissions	Keep the default Ask for permission at first. Inspect the target file or command before approving edits or higher-risk commands. You can change any session separately with the composer control; see the five modes.	Default for new chat sessions; change an existing session in its composer.
Response language	Choose a language if you want consistent replies in it. This is independent of the interface language and is unset by default.	Subsequent replies; earlier messages are unchanged.
Output style	Default is concise; Explanatory explains implementation choices; Learning asks you to write small pieces. Installed custom styles can also appear here.	New sessions, sessions resumed in a new process, or restarted sessions. A running session keeps its current prompt.
Effort level and Thinking Mode	Start with the supported model default. Adjust for speed, cost, or provider compatibility; disabling thinking sends a non-thinking parameter to compatible providers that require one.	Defaults for new sessions; available effort levels depend on the selected model.

Try it: Choose a disposable Git project and ask: “List the files you plan to change, make one small edit, then tell me how to verify it.” Inspect any permission prompt, then review the Diff file by file.

Appearance and input

• Color theme: Pure White, Paper, Warm Classic, Celadon, Ink Night, or Ink Blue. Follow the system lets you choose separate light and dark themes.
• Language: changes only the interface, with English, Simplified and Traditional Chinese, Japanese, and Korean choices. It does not set the response language.
• UI Zoom: scales the whole window. Use ⌘+ / ⌘- on macOS or Ctrl+ / Ctrl- on Windows; press 0 to reset to 100%.
• Message sending: defaults to Enter to send and Shift+Enter for a newline. If you write long prompts, choose Ctrl/Cmd+Enter to send.
• Default editor: chooses which detected local editor appears as the default under a file's Open with menu.

For more involved agent workflows

Setting	When to use it
Ultracode	On by default. A standalone keyword triggers dynamic Workflow orchestration; the same text in code blocks, quotes, or paths keeps its literal meaning.
Agent Teams	On by default. Use it when a larger task needs several agents working together. New sessions pick it up; existing sessions do after an app restart. See Subagents first.
Auto-answer questions	Off by default. After 1, 5, 10, or 30 minutes without an answer to an agent's multiple-choice question, it can choose a recommended option. If it cannot choose reliably, it keeps waiting. Use only when you want unattended work to continue.
Auto-dream	Off by default. After enough sessions have accumulated, it can organize auto-memory in the background, using additional model calls and tokens.
Agent Trace	On by default. New sessions write raw request, response, and status events to a local traces directory. To investigate a failure, open an assistant line in the session's Trajectory and switch to its Raw request tab; turning it off stops new records, while old ones remain readable.

Network, web search, and notifications

• System Notifications are off by default. Enable them and grant OS permission when you need alerts for permission requests, completed replies, or scheduled tasks away from the app.
• Network uses the system proxy by default. Direct bypasses system and inherited process proxies; Manual accepts an HTTP/HTTPS URL such as http://user:password@127.0.0.1:7890. Click Save. New requests use the new route; requests already in flight keep their old route. Model services, account sign-in, MCP, and agent tools use this setting. App updates have a separate proxy under About.
• AI request timeout defaults to 1800 seconds and accepts at least 30 seconds. Raise it only when the first model response or connection test really times out. For a local model that thinks for a long time, you can enter 14400 seconds (four hours).
• WebFetch preflight skips the upstream domain check by default to avoid false failures with third-party providers or restricted networks. Turn this off only when you specifically want the upstream check.
• WebSearch defaults to Auto: Claude models try native search first, then Tavily or Brave on failure or with non-Claude models. You can force Claude, Tavily, Brave, or Off. Tavily and Brave require your own API keys and a click on Save.

Local data: check the effect before changing it

• Session retention defaults to 365 days and accepts 0–3650 days. Reducing it immediately removes older records; setting 0 removes all existing records and stops recording session contents. The UI previews and confirms the change. This is not a backup system.
• Data Storage Location defaults to ~/.claude (%USERPROFILE%\\.claude on Windows). Portable mode takes an absolute path outside the app installation folder. After an app restart, sessions, configuration, tasks, skills, and plugins are read from the new location. The two directories are not merged or migrated automatically. If CLAUDE_CONFIG_DIR is set at launch, remove that environment variable before switching locations in the UI.

Product screenshots in this guide consistently use the Pure White theme so the interface can be compared without palette changes.

H5 Access

Continue the same session in your phone's browser. Off by default. See Phone (H5) and IM.

IM Adapters

Talk to Claude from WeChat, DingTalk, WhatsApp, Telegram, Feishu, WeCom, QQ, or Slack, and manage paired users. See Phone (H5) and IM and IM integrations.

Terminal

A real host shell embedded in the app, for installing plugins, skills, MCP servers, and anything else that needs a command line. The desktop app bundles claude-haha, so anywhere the docs say claude <args> you can run claude-haha <args>.

On Windows you can also choose the startup shell (system default, PowerShell 7, Windows PowerShell, Command Prompt, or a custom executable) and set a Bash path — used when a tool calls Unix commands like grep or sed, usually pointing at Git Bash.

MCP

External tools and data sources. STDIO, Streamable HTTP, and SSE transports are supported, and the scopes match the CLI:

• Local — only for you, but bound to one project.
• Project — written to the project's .mcp.json and shared with the team.
• User — written to your global config, active in every project.

The three numbers at the top are total servers, currently connected, and needs attention. STDIO commands run directly on your machine, so runtimes like Node, Python, and Bun must be installed and on your PATH.

Agents

Browse installed agents and create your own. See Subagents.

Skills

Every skill available on this machine, grouped by source, with the prose and source files readable in place. See Skills and the Skills Market.

Memory

View and edit the Markdown memory files Claude keeps per project. Pick a project on the left, a file in the middle, then edit or preview the rendered output. The files live in ~/.claude/projects/<project>/memory/ and are loaded by the CLI at runtime.

/memory in any session jumps straight here. For how memory is written and recalled, see Memory system.

Plugins

A plugin bundles skills, agents, hooks, and MCP servers together. This tab shows installed plugins, their health, and what capabilities each one exposes, with enable, disable, update, and uninstall — including multi-select for bulk operations.

After enabling or disabling anything, click Apply changes to push the change into the running runtime.

Pets

A little robot floating on your desktop. Off by default. See Desktop pet.

Computer Use

Let Claude read the screen, click, and type. Enable it, confirm the global consent, and satisfy the OS permission checks; setup steps vary by platform. See Computer Use.

Token usage

A usage dashboard computed from the Claude Code session records on this machine. Everything is calculated locally and nothing is uploaded.

• Across the top: total tokens, peak tokens, longest task, current and longest streak.
• In the middle: a heatmap with daily, weekly, and cumulative views. Click a day for that day's sessions, tokens, messages, and tool calls.
• Below: activity insights — active rate, most-used model, skills used, fresh versus cache-hit token split, and estimated cost. Estimated cost excludes models with no known price, and the UI says how many were skipped.

Diagnostics

Where to go when something breaks. Logs server and CLI startup, provider, and session runtime errors.

• Across the top: log size, event count, warnings in the last 24 hours, retention policy.
• Recent events lists the actual errors, each with an event ID you can copy on its own.
• Export Bundle / Copy error summary / Copy issue report — use the latter two when filing an issue; they're already formatted.
• Doctor — checks user and current-project configuration state, read-only, returning a healthy / not configured / missing / invalid list. /doctor in a session opens it directly.
• Reset safe UI state — clears only regenerable interface keys like open tabs, theme, and zoom. Chat history, model config, skills, MCP, IM, and OAuth are always protected and never touched.
• Local index — status and size of the derived SQLite index, with a rebuild button. Rebuilding only affects the index; source conversations are never deleted.

:::info Issue reports and exported bundles are redacted on a best-effort basis — chat contents, file contents, full environment variables, and API keys are omitted. Still give them a read before sharing, in case an internal hostname, username, or path slipped through. :::

About

Version, changelog, GitHub repo, feedback link.

App Updates checks GitHub Releases, downloads, and restarts to install. Updates use their own proxy setting, separate from Settings → General — if updates stall on a corporate network, configure the advanced update proxy here.