Every setting

Every section of Settings and what is in it, in the order the app shows them, with a link to each and to the part of this guide that explains it. Use it to find where something is set; the topics explain how to use it.

How Settings is laid out#

Open Settings from the sidebar. On a computer, Settings has three tabs at the top: App (the app itself), Agents (the agents and what they read) and Capabilities (what the app adds to every agent: search, the browser, memory and the rest). Each tab lists its sections on the left, and the open section fills the page. On a phone, Settings lists every section in one list; tap one to open it, and the arrow at the top to go back.

You don’t have to know where an option lives: type what you want in the search box (Ctrl+K, ⌘K on a Mac), such as "dark mode" or "notifications", and pick it. The app opens the right section and points at the option. You can also ask any agent to change a setting for you (Let agents work the app for you).

App#

General#

Settings → General: theme, and what Enter does while an agent works. More in Look, language and updates.

• Theme: Dark, light, or following your device; or a theme of your own.
• New theme: Make a theme of your own from any of them.
• Import: Add a theme someone shared.
• Language: The language the app is shown in, and agents write in.
• Sending while an agent works: Enter joins the running turn, or waits for it to end.
• Terminal suggestions: Grey completions after the cursor in terminals.
• Resume after usage limits: Continues a session a minute after a plan’s limit resets.
• Delete inactive sessions: Deletes sessions nobody used for a year, or as long as you choose; until then every conversation is kept.
• Start when this computer starts: Starts the app when the computer starts.
• Update automatically: Installs new versions by itself: the desktop apps as soon as they are out, after a minute’s notice; a copy from git when you have not used it for a while.
• Take the tour: See what each part of the app does.
• Tips: Points out a part of the app the first time you come to it, after the welcome tour.
• Check now: Look for a newer version now.

Notifications#

Settings → Notifications: when a session finishes, fails or needs your answer. More in Inbox and notifications.

• Notify this device: A notice when a session finishes, fails or needs your answer, even with the app closed.
• Sound: A short chime with each notice while the app is open.
• Send a test: Send a notice to this device now.

Password#

Settings → Password: the password for this app, and signing out. More in Look, language and updates.

• Change password: Changing it signs out every other browser.
• Sign out of this browser: You need the password to come back.

Source control#

Settings → Source control: GitHub and Gitea accounts for opening pull requests. More in Phones, other machines and sharing.

• Add an account: An account to open pull requests with.

Connect other agents#

Settings → Connect other agents: let agents elsewhere start and steer sessions here, through this app’s MCP server. More in Phones, other machines and sharing.

• New token: A token for an agent elsewhere to start and steer sessions here.
• What they can do: The tools an agent with a token gets.

Machines#

Settings → Machines: other machines running Hydor, to use from here, and access to this one. More in Phones, other machines and sharing.

• Add a machine: Another machine running Hydor, to use from here.
• Access from other machines: Tokens other machines use to reach this one.

Agents#

Accounts#

Settings → Accounts: the Claude Code and Codex sign-ins every session uses. More in Agents, accounts and models.

• Your accounts: The Claude Code and Codex sign-ins sessions use, and more you add.
• Switch accounts at a limit: A session that runs out of usage carries on with another account.
• Agent programs: Which agent programs are installed, and installing the others.
• Update automatically: Keeps Claude Code, Codex and the other agent programs at their newest version.

More agents#

Settings → More agents: Gemini CLI, OpenCode, Qwen Code, Goose and other agents that speak ACP. More in Agents, accounts and models.

• Your agents: Agents that show next to Claude Code and Codex.
• Add your own: Any agent that speaks ACP, by its command.
• Agents that speak ACP: Gemini CLI, OpenCode, Qwen Code and Goose, ready to add.

Models#

Settings → Models: models for Custom, the app’s own agent: OpenRouter, Ollama, and Hugging Face models run here with vLLM. More in Agents, accounts and models.

• Helper model: The model for small jobs beside sessions.
• On this machine: Models installed here, and their downloads.
• Add a model from Hugging Face: The ones suggested for this computer, or search, see how each would run here, and install it.
• OpenRouter: Models through OpenRouter, with your key.
• Runtimes and endpoints: Ollama, vLLM and other OpenAI-compatible endpoints.
• Hugging Face token: Only for gated or private models.

Agent Guidelines#

Settings → Agent Guidelines: what every agent follows: in every session, in a project, or in one session. More in Agent Guidelines.

• For every session: What every agent follows, in every session.
• For a project: Guidelines the sessions in a project follow.
• For one session: Guidelines for one session, from the start page or the session’s menu.
• Files Claude Code and Codex read too: CLAUDE.md and AGENTS.md, which Claude Code and Codex read on their own.

MCP servers#

Settings → MCP servers: tools the agents can call. More in Connected apps, keys and tools.

• Add server: A tool server for Claude Code or Codex.
• Check connections: See whether each MCP server answers.

Skills#

Settings → Skills: written methods agents load when a task calls for one: built in, yours, from plugins and the agents’ own. More in Skills.

• New skill: Write a skill for every agent, or for Claude Code or Codex alone.
• Built in: The skills that come with the app.
• Yours: Skills you wrote, or approved when an agent suggested them.
• From plugins: Skills your plugins bring.
• Claude Code and Codex: The agents’ own skills, which only they get.
• Suggest skills that fit: A new session’s agent hears which skills fit its first message.
• Bring skills in by keywords: Keywords in a message load skills by themselves.
• Make my own version: Change a built-in skill: your version takes its place.

Plugins#

Settings → Plugins: add-ons that bring skills, tools and instructions to every agent. More in Connected apps, keys and tools.

• Your plugins: The plugins you added, on or off for every agent.
• Come with the app: Plugins that come with the app, ready to add.
• Make a plugin: Bundle skills, tools, instructions and files of your own as a plugin.
• From a link: Add a plugin from GitHub or another Git address.
• From a .zip file: Add a plugin someone packed into a .zip.
• Pack into a .zip: A plugin as one file to share.
• Claude Code: Claude Code’s own plugins and marketplaces.
• Codex: Codex’s own plugins and marketplaces.

Scheduled tasks#

Settings → Scheduled tasks: agent work that starts by itself: on a schedule, once, or when a webhook calls. More in Scheduled tasks.

• New task: Agent work that starts by itself.
• Start from an example: Ready-made scheduled tasks.
• Morning brief: A ready-made scheduled task.
• News on a topic: A ready-made scheduled task.
• Price watch: A ready-made scheduled task.
• Page change alert: A ready-made scheduled task.
• Weekly report from a folder: A ready-made scheduled task.
• Inbox digest: A ready-made scheduled task.
• Habit check-in: A ready-made scheduled task.
• New releases: A ready-made scheduled task.
• Site up check: A ready-made scheduled task.
• Disk space alert: A ready-made scheduled task.
• Review new pull requests: A ready-made scheduled task.
• New files in a folder: A ready-made scheduled task.
• Reminder: A ready-made scheduled task.

Capabilities#

Search#

Settings → Search: how agents search the web: Naget Search and other providers. More in Search and the browser.

• Providers: Which search providers agents use, and in what order.
• Naget Search: The app’s own web search.
• SearXNG: A search engine you run yourself.
• Brave Search: Web search with a Brave key.
• DuckDuckGo: Web search without a key.
• Results per search: How many results a search gives.

Browser#

Settings → Browser: a web browser agents use, which you can watch and take over. More in Search and the browser.

• A browser for agents: Agents open sites, click, type and read pages.
• Chromium: Which Chromium the browser runs.
• Show a window: Run the browser in a window you can see.
• Keep pages off this machine and its network: Sites on this computer and its network stay off limits.
• Screenshots for models that see images: Models that see images get screenshots of the page.
• Sites: Only these sites, or never these.
• Profiles: Separate sign-ins for the browser.

Computer use#

Settings → Computer use: agents see your Mac’s screen and use its mouse and keyboard, in the Mac app. More in Search and the browser.

• Let agents use this Mac: Agents see the screen and use the mouse and keyboard, in any app; each session asks you first.
• Accessibility: macOS’s permission for Hydor to use the mouse and keyboard.
• Screen Recording: macOS’s permission for Hydor to see the screen.

Memory#

Settings → Memory: what agents remember about you, your projects and each session, and when it was written. More in Memory.

• Memory: Agents start with what earlier sessions kept, and save notes when you ask.
• Search by meaning: Finds notes, and anything in this search, by what it means.
• Notes each agent starts with: Besides your profile: the notes that bear most on what you ask, then the latest.
• Tidy up each night: The helper model merges notes that say the same thing and sums related ones up.
• Tidy up now: Tidies memory up at once.
• Your profile: What every agent gets about you at the start of a session; agents keep it current.
• Profile length: The most words your profile may have; longer, the helper model shortens it.
• Notes: Every note kept, by place.
• Add a note: Write a note for agents to remember, about you or a project.
• Deleted notes: Notes deleted in the last 30 days, to restore.
• Export: Every note as Markdown to read, or as JSON with histories.
• Clear all…: Deletes every note and its history.

Child sessions#

Settings → Child sessions: agents handing parts of a job to other agents, and swarms for hard problems. More in Deep research, second opinions and several agents.

• Working at once, per session: How many child sessions one session runs at once.
• Swarm size: How many agents of a swarm work at once.

Images, video and audio#

Settings → Images, video and audio: pictures, videos, music and voices agents make for you through OpenRouter, and who writes their captions in the Library. More in Documents, images, videos and voice.

• Making images: Let agents make images.
• Image model: The model that makes them.
• Making videos: Let agents make videos.
• Video model: The model that makes videos.
• Making audio: Let agents make audio.
• Music model: The model that makes music.
• Voice and sound model: The model that makes voices and sounds.
• Captions in the Library: The model that writes the titles and descriptions of the images, videos and audio you keep.

Voice#

Settings → Voice: dictate messages, and have replies read aloud. More in Documents, images, videos and voice.

• Dictation: Speak a message instead of typing it.
• Language you speak: The language dictation listens for.
• Speech recognition on the server: Recognize speech on this computer or with a service.
• Read aloud: Have replies read aloud.

Assistants#

Settings → Assistants: presets to start sessions as (Researcher, Writer, Tutor…), and where task folders go. More in Assistants.

• New assistant: A preset to start sessions as.
• Clean up after a task: Agents remove what they made along the way and no longer need.
• Task folders: Where an assistant’s new task folders go.

Custom#

Settings → Custom: auto mode, a sandbox for commands, and how long goals run. More in Custom, the app's own agent.

• Auto mode: Custom approves safe steps by itself.
• Sandbox for commands: Commands run in a sandbox.
• Goals: How many rounds a goal may run.

Ask an agent#

• "Where do I change what Enter does while an agent works?"
• "Turn on the notification sound."
• "Show me my search settings."
• "Set the helper model to the cheapest one."
• "Turn off memory for new sessions."