Basic Tab

The Basic tab is where you shape who the agent is and what it's trying to do. It's the highest-leverage tab in the product — most quality issues are fixed here, not in the LLM or voice tabs.

Agent name

The display name shown across the dashboard, call logs, and the agent's self-introduction (when greetings are enabled). Change this any time — it doesn't affect the agent ID.

First message (greeting)

A short message the agent speaks the moment a call connects, before the user says anything. Useful for inbound calls where you want the agent to identify itself immediately.

• Toggle off if you want the agent to wait for the caller to speak first (common for outbound calls where you're picking up mid-ring).
• Toggle on and write something like: "Hi, this is Aria from Acme Support. How can I help you today?"

When greetings are enabled, the message text becomes part of the agent's first turn. Keep it short — 1–2 sentences max.

Persona

A high-level archetype that pre-tunes the system prompt. Pick the one closest to your use case:

Persona	Best for
Helpdesk	Calm, patient support; troubleshooting; FAQs
Sales	Confident, consultative outbound; lead qualification; upsells
Healthcare	Empathetic intake; appointment confirmation; sensitive conversations

The persona affects tone words, default conversation pace, and small-talk behavior in the generated system prompt. You can always edit the prompt directly afterward.

Voice gender

Male / Female. This biases voice selection on the Voice tab and is referenced in the system prompt (so the agent uses the right pronouns when talking about itself).

Primary language

The language the agent speaks and listens for. Picking a language here also updates the Speech tab — Bolti switches the STT model and language code so transcription works correctly out of the box.

Supported languages include English (multiple regions), Hindi, Spanish, French, German, Italian, Portuguese, Arabic, Japanese, Korean, Chinese, and many more. The dropdown shows everything the current STT providers support.

Timezone

Used by the agent for any time-of-day reference (e.g. "good morning", "we're open until 6 PM") and for scheduling features. Defaults to Asia/Kolkata.

Goal

A single sentence that describes the outcome you want from each call. Examples:

• "Qualify the lead by collecting budget, timeline, and decision-maker, then book a demo."
• "Help the customer reset their password and verify their account is accessible."
• "Confirm the patient's upcoming appointment and reschedule if requested."

The goal is the most important free-text field on this tab. A specific, action-oriented goal produces dramatically better calls than a vague one. Aim for one sentence that names the outcome and the information the agent should collect.

Guardrails — forbidden topics

A list of subjects the agent must refuse to discuss. Each guardrail is a short phrase like:

• pricing for enterprise plans
• legal advice
• competitor product comparisons
• personal opinions on politics

Type a topic, hit Enter (or click +), and it's added. The system prompt instructs the agent to politely decline and steer the conversation back when any of these come up.

Company domain (optional)

Enter your company's website (e.g. acme.com) and click Fetch Info. Bolti scrapes the homepage, extracts a concise company-context paragraph, and injects it into the system prompt. The agent will know what your company does, your products, and your tone — without you having to write any of it.

You can edit or clear the fetched context after it loads.

When to use this

Use the company-domain fetch for any agent that represents your business externally. Skip it for purely internal use cases (e.g. an internal IT helpdesk that doesn't need to talk about the company).

Smart Call Scheduling

Toggle this on to enable intent detection for callback requests. When a caller says something like "can you call me in 10 minutes" or "reach out tomorrow at 3", Bolti automatically:

1. Parses the requested time using the agent's timezone.
2. Registers an outbound task in the Schedule tab on the agent's detail page.
3. Dials the same caller back at the requested time.

The agent will confirm the scheduled time on the call before hanging up.

System prompt

The full prompt sent to the LLM on every turn. By default, this is auto-generated from your wizard answers above — every time you change a wizard field, the prompt is recompiled.

Click Show system prompt to view the generated prompt. You can:

• Read it to understand exactly what the agent is being told.
• Edit it directly to override the auto-generated version. Once you edit, your changes are kept and not overwritten by future wizard changes.
• Reset to generated to discard your manual edits and regenerate from the wizard.

When the prompt matches the auto-generated version, an Auto-generated badge appears in the textarea. When you've manually edited it, the badge disappears and your prompt is the source of truth.

When to edit the prompt manually

• The wizard fields don't capture a constraint you need (e.g. "never use the word 'unfortunately'").
• You're porting a prompt from another platform.
• You want full control over the agent's voice and the wizard isn't expressive enough.

For everything else, prefer adjusting the wizard fields — they keep the agent's behavior consistent and easy for teammates to understand.

Dynamic variables

Both the System prompt and the First message support dynamic variables — anything you write inside {{ ... }} is substituted at call time, so one agent can serve many callers, contexts, or campaigns.

Hello {{ name | default: "there" }}, your appointment is on {{ appointment_date }}.

Variables without a default: filter are required and must be supplied by the caller (via the API or the preview dialog); built-ins like {{ customer.number }}, {{ date }}, and {{ year }} are filled in automatically.

See the dedicated Context & Dynamic Variables page for the full reference: built-in variables, validation rules, payload limits, the discovery endpoint, and end-to-end API examples.

Saving

Like every tab, your changes apply once you click Save Changes at the top of the Settings panel. Until then you can switch tabs freely without losing edits.