Quick Start
This guide walks you from zero to a voice agent you can call and talk to in roughly five minutes. We'll create an account, build an agent with the wizard, test it in your browser, and (optionally) attach a real phone number.
What you'll need
- A working email address (for signup)
- A modern browser with microphone access (for the in-browser preview)
- A phone number you can dial out to (only if you want to test outbound calling)
You do not need any API keys, telephony account, or LLM credits to complete this guide — Bolti's defaults work out of the box.
1. Sign up
Open the Bolti dashboard and create an account:
You can sign in with email/password or with Google. On first login, Bolti automatically creates an organization and a default workspace for you. Every agent, phone number, and tool you create lives inside a workspace.
An organization is your top-level account (your team or company). A workspace is a sandbox inside that org — most teams use one workspace per environment (e.g. production, staging) or per customer if you're building for clients. See Workspaces & Organizations for the full model.
2. Create your first agent
From the dashboard home (the Assistants page), click Create AI Agent in the top right.
A wizard opens and walks you through six short steps:
| Step | What you're choosing | Defaults |
|---|---|---|
| Agent name | A friendly identifier (e.g. "Acme Support") | — |
| Persona | Helpdesk, Sales, or Healthcare — shapes tone and behavior | Helpdesk |
| Voice gender | Male or Female | Female |
| Language & timezone | Primary conversation language and timezone for time references | English / Asia/Kolkata |
| Goal | One sentence describing what the agent should accomplish on each call | — |
| Voice | Pick a voice from the curated voice cards (you can preview each one) | Auto-selected based on language + gender |
You can also (optionally):
- Enter a company domain like
acme.comand click Fetch Info — Bolti pulls your website and writes a company-context paragraph the agent uses on calls. - Add guardrails — topics the agent must refuse to discuss.
- Toggle Smart Call Scheduling — the agent will detect phrases like "call me back in 10 minutes" and queue an outbound follow-up automatically.
When you click Create, Bolti compiles your wizard answers into a system prompt, applies sensible defaults for LLM/STT/TTS, and drops you straight into the agent's settings page.
Every wizard answer is editable from the Basic tab on the agent's settings page. Editing any field automatically regenerates the system prompt — or you can take over the prompt manually.
3. Test it in your browser
On the agent detail page, click the Preview tab.
This opens a built-in browser call client. Hit Connect, allow microphone access, and start talking. You're talking to your agent over the same realtime stack that powers production calls — no phone number needed.
Use this loop while you iterate on the prompt: tweak the goal or guardrails on the Basic tab, click Save Changes, then re-connect in Preview to hear the new behavior.
4. (Optional) Make a real phone call
If you want to dial a real number, you need an outbound-capable phone number on the workspace.
- Go to Phone Numbers in the left nav.
- Either Buy a number (Bolti provisions one for you) or bring your own by registering a SIP trunk — see Telephony → Connect a Provider.
- Back on the agent detail page, click the Make Outbound Call button at the top right.
- Pick the outbound number, enter the destination phone number in E.164 format (e.g.
+919876543210), and dial.
Bolti queues the call, dials out via your trunk, and the agent runs the conversation. The call will show up in the Logs tab once it completes, with full transcript and recording.
5. (Optional) Receive inbound calls
To have your agent answer incoming calls:
- Go to the agent's Settings → Phone tab.
- Pick a SIP trunk and the inbound phone number you want to assign.
- Click Assign Number.
That number now routes any incoming call to this agent. See Receive Inbound Calls for advanced routing (IVR, caller matching, etc.).
What's next
Now that you have a working agent, here's what to explore next based on what you're building:
- Tweaking how it sounds and feels → Agent Setup overview
- Connecting your own systems (CRM, calendars, APIs) → Tool Calling
- Letting the agent answer questions from your docs → Knowledge Bases
- Calling lots of people on a schedule → Batch Calling
- Driving Bolti from Cursor or Claude Desktop → Bolti MCP Server
If you get stuck, check the FAQ or browse the full API Reference.