Brain & Settings
You configure an agent across two tabs. The Brain tab shapes how it thinks and responds; the Settings tab covers customer access, groups, and deployment. (Its name, description, and type are set when you create it.)
The Brain tab
The Brain tab holds the agent's AI Config, along with its Knowledge, Data, and Memory — the knowledge bases and data tables it draws on, and what it remembers over time.
AI Config controls how it responds:
| Setting | What it does |
|---|---|
| Temperature | Low = precise and consistent; high = more creative and varied |
| Response length | Concise, Balanced, or Detailed |
| Max tokens | A cap on how long a single response can run |
| System prompt | The core instruction that defines who the agent is and what it does |
| Supplemental instructions | Extra rules appended to the system prompt |
| Brand guidelines | Strict / Casual / Inherit — the safety and content rules themselves are managed at the brand level (Brand Safety), so they apply across all of a brand's agents |
Escalation — define when the agent hands off to a human: the triggers (keywords or topics like "refund" or "legal") and the behavior (Offer to Connect, Auto Escalate, or Provide Contact Details).
The Settings tab
- Deploy — publish or unpublish, copy the landing-page URL and widget embed code, and map a custom domain. → Deploying Your Agent
- Access — control who can open the agent: Public (anyone with the link), Domain Restricted (specific email domains), Email Allowlist, or Password Protected (SSO is coming soon). Internal agents get a nudge to restrict access.
- Customer Auth — verify a visitor's identity before they chat (see below).
- Groups — add the agent to an Agent Group to share context with other agents.
- Dynamic Page — adapt the landing page to each visitor: any URL parameter becomes a
{{url.key}}variable, and you can feed it from a linked product catalog. (Visitor location, campaign/UTM tracking, and time-of-day are coming soon.)
General holds the agent's name, description, and type; Danger is where you delete the agent.
Customer Authentication
Customer Authentication lets a conversational agent verify who it's talking to before the conversation starts, so it can gate sensitive information. (This is separate from team access — it identifies the end-users chatting with your agent, not who can manage it.)
Under Customer Auth:
- Enable customer login — require visitors to verify their identity before chatting.
- Method — Email OTP (the visitor enters their email and gets a one-time code) is available today; Google sign-in and Phone OTP are coming soon.
- Login card branding — set the title and subtitle shown on the sign-in card.
- Session duration — how long a visitor stays signed in before they need to verify again (for example, 24 hours).
You can also require sign-in at a specific point in a conversation by adding a Login step in the Journey Builder. Auto-fetching a signed-in visitor's profile from your CRM or API is coming soon.