Skip to main content
Tinfoil Chat separates global preferences from controls that apply to one conversation. Global preferences are available from Settings; model, prompt, and Web Search controls can also appear in the message composer.
Open the sidebar and select the Settings cog. Chat behavior is primarily under Chat Settings, Personalization, and Prompts.
Web Search lets the model retrieve current information while answering. It is on by default. There are two controls:
  • The global Web Search setting makes the feature available and sets the initial value for new chats.
  • The per-chat Web Search control decides whether the current conversation can search.
Turning the global setting off forces Web Search off and hides its per-chat control. It does not erase the preference stored on an existing chat, so that preference is available again if you re-enable the global setting.
  • Global setting: Settings → Chat Settings → Advanced Settings → Web Search
  • Current chat: open Input options in the composer and use Web search
Automatic PII Blocking in Web Search is also under Advanced Settings.
When Automatic PII Blocking is enabled, web-search requests include an automatic PII check. It is on by default on both platforms.

Generative UI

Generative UI (GenUI) allows the assistant to return interactive widgets, such as charts and timelines, instead of text alone. It is on by default and applies globally rather than per chat.
Go to Settings → Chat Settings → Advanced Settings → Generative UI.
When enabled, new requests can use interactive widgets. When disabled, the app does not send those tool capabilities. Disabling Generative UI does not remove widgets already saved in conversations.
Some models do not support tool use. When you use Auto, Tinfoil Chat prefers a compatible model when Generative UI is enabled and one is available.

Reasoning and thinking

Reasoning controls are model-dependent rather than a Settings row.
Open the model selector. Supported models show Effort with Low, Medium, and High and, when supported, a Thinking switch.
Medium effort and enabled thinking are the defaults. Models that do not advertise a reasoning capability do not show these controls. Higher effort can produce a more thorough response but usually takes longer. The reasoning preference is global, but Tinfoil Chat applies it only when the selected model supports it.
On iOS, Auto tiers do not expose reasoning controls because they do not represent one model with a fixed reasoning configuration.

Personalization

Personalization adds information about you to the model context.
Go to Settings → Personalization. You can set Name, Occupation, Conversational Traits, and Additional Context, then select Save.Use Personalize responses to turn profile insertion on or off. Turning it off keeps the saved fields but stops adding them to new requests.
Personalization is on by default.

System prompts and prompt presets

A system prompt changes how the assistant behaves. You can set a global default or select a preset for one chat. The precedence is:
  1. The prompt preset selected for the current chat
  2. Your custom default system prompt
  3. Tinfoil Chat’s built-in default
  • Settings → Prompts → Default System Prompt sets the custom default.
  • Built-in Prompts and Your Prompts manage the prompt library.
  • In the composer, open Input options → Change system prompt to select a preset for the current chat.
Custom default prompts and prompt presets support {USER_PREFERENCES}, {LANGUAGE}, and {TIMEZONE} placeholders.

Language, appearance, and chat font

  • Settings → General → Appearance → Theme: Light, Dark, or System
  • Settings → Chat Settings → Conversation Settings → Chat font: System, Serif, Mono, or Dyslexic friendly
  • Response Language in Conversation Settings controls the language requested from the model.
Both platforms provide System, which resolves the current browser or device language when a request is sent. The literal word System is never sent to the model.

Model selection

Model selection is not a Settings row. Use the model button in the message composer to select a model or an Auto tier. Tinfoil Chat remembers a device-local fallback model for new chats. Saved web chats retain their selected model. On iOS, downloaded chats use the model selected on that device. When available, the initial default is Auto · Fast.

What syncs across devices

With cloud sync enabled, Tinfoil Chat encrypts and syncs these profile settings:
  • Web Search availability and automatic PII checking
  • Generative UI
  • Reasoning and thinking preferences
  • Language
  • Personalization
  • Custom prompts, prompt presets, and favorites
  • Web theme and chat font
The active prompt preset and Web Search preference are stored with the chat. Web also stores the selected model with the chat; iOS cloud sync does not. The fallback model for new chats and the Enable local chats preference remain device-local. Settings available only on web remain unchanged when you use iOS. See cloud sync and encryption for how settings are protected, and data controls and deletion for local and temporary chat options.