Skip to main content

Personalization

Use Personalization to tell agents how to address you and which standing preferences they should consider during chats.

Personalization applies across agents. It is different from conversation history, which belongs to a specific chat thread, and different from connectors, which authorize external tools.

Open the ☰ drawer and select Personalization.

The page has three stacked sections:

  • Name Display — your account name, optional nickname, and name-use toggle.
  • My Preferences — preferences you add, edit, or delete manually.
  • Auto-generated by AI — preferences the AI may learn from your usage over time.

Name Display​

The Name Display section controls how agents refer to you.

It shows:

  • Your name — the account name known to the platform, such as your username or registered display name.
  • Nickname — an optional name you want the AI to use instead.
  • Let the AI use my name — a toggle that controls whether agents can use your name in replies.

The nickname field is optional. If you leave it blank, the platform can still show your account name in this section.

The name-use toggle is on by default. When it is on, agents may address you by your nickname or account name. When it is off, the name remains saved in your profile settings, but it is not provided to agents as part of personalization.

My Preferences​

My Preferences is a list of short, free-form instructions or context that agents can use in future chats.

Good preferences are specific and reusable:

  • "I prefer concise answers."
  • "Reply in French."
  • "I'm a Python developer; explain things using Python analogies."
  • "I'm based in Berlin."

Avoid adding one-time requests here. If a preference only applies to the current task, send it in the chat instead.

Add a preference​

  1. In My Preferences, click Add.
  2. The Add Preference dialog opens.
  3. Type the preference in the Preference field.
  4. Click Add.

The dialog includes the helper text "Tell the AI something about your preferences" and an example placeholder such as "I prefer concise responses".

The new preference appears in the list immediately.

Edit a preference​

  1. Click the pencil icon on the preference row.
  2. The Edit Preference dialog opens with the current text already filled in.
  3. Update the text.
  4. Click Save.

Use edit when a preference is too vague, outdated, or creating unwanted behavior.

Delete a preference​

  1. Click the trash icon on the preference row.
  2. The Delete Preference confirmation dialog opens.
  3. Review the quoted preference text in the dialog.
  4. Click Delete to remove it, or Cancel to keep it.

Deleting a preference removes it from future agent context. There is no per-preference on/off toggle; a preference is either present or deleted.

Auto-generated by AI​

The Auto-generated by AI section appears below My Preferences.

It is intended for preferences the AI learns from your interactions over time. The page may show placeholder text such as:

As you chat, the AI will learn your preferences and they will appear here.

Manually added preferences remain the most direct way to control personalization. If you want agents to consistently follow a preference, add it under My Preferences instead of waiting for it to be inferred.

How agents use personalization​

When you chat with an agent, the platform can include your personalization settings in the agent's context. This gives the agent standing guidance, such as:

  • What to call you.
  • Which language or tone you prefer.
  • How much detail you want.
  • Background context that should shape answers.

For developers and agent builders, treat personalization as user-scoped context. It is not a hard permission system and it does not replace tool configuration, system prompts, or connector authorization. Agents still need to follow their own instructions and tool policies.

If an agent ignores a preference, make the preference more specific or restate the requirement in the chat.

What's not personalization​

  • Conversation memory — previous messages belong to a specific conversation. See Conversations.
  • Connector tokens — external service authorization is managed on the Connectors page.
  • Profile details — account details such as email and password are managed on Profile.
  • Agent configuration — model choice, tools, and system behavior are configured by agent creators or admins.

Developer notes​

  • Personalization is user-scoped. A preference added by one user should not affect another user.
  • Preferences are meant to be reusable context, not one-off prompts.
  • Manual preferences are explicit user-provided data and should be treated as higher-confidence than inferred preferences.
  • Deleting a preference should remove it from future agent context.
  • Name usage should respect the Let the AI use my name toggle.

Tips​

  • Be specific. "Reply in 2-3 sentences" beats "be brief".
  • Add context, not commands. "I'm a designer, not a developer" gives agents useful framing for every reply.
  • Keep the list short. Too many competing preferences make it hard for the agent to satisfy all of them.
  • Iterate. Add a preference, see how an agent behaves, refine. Two minutes of personalization saves repeating yourself in every chat.

On this page