Skip to main content
New to Bolna? Start with What Bolna does. This page explains what happens under the hood: the voice pipeline, the terms you will see across the docs, what an agent can do, and how an agent is configured.

​
How voice AI works

Every conversation flows through a three-step pipeline that happens in real-time:

Listen

Speech-to-Text (ASR) converts the caller’s voice into text that your AI can understand.Bolna supports Deepgram, Azure, ElevenLabs, and more.

Think

Large Language Model (LLM) processes the text, understands context, and generates an intelligent response.Connect OpenAI, Anthropic, Azure OpenAI, or your own LLM.

Speak

Text-to-Speech (TTS) converts the response into natural-sounding speech played back to the caller.Choose voices from ElevenLabs, Cartesia, Azure, and more.
Bolna orchestrates this entire pipeline in under 600ms, enabling natural, real-time conversations with minimal latency.
Agents can also run on a single speech-to-speech model that hears and speaks directly, without the separate transcribe and synthesize steps. See speech-to-speech agents.

​
Key terms

New to voice AI? The Glossary explains every term used across these docs in plain English, including LLM, transcriber (ASR), synthesizer (TTS), telephony provider, prompt, latency and knowledge base (RAG).

​
What can Bolna agents do?

​
Handle phone calls

Inbound calls

Answer customer calls 24/7 with AI-powered responses

Outbound calls

Proactively reach customers for sales, reminders, and follow-ups

Batch calling

Launch campaigns with thousands of concurrent calls

Call transfer

Route to human agents when AI assistance isn’t enough

​
Execute actions during calls

Agents can perform real-time actions by calling external APIs and tools:
Configure function tools in the Tools Tab. Your agent can access CRMs, databases, payment systems, and more, all while on the call.

​
Extract insights after calls

Every conversation generates valuable data:

Transcripts

Full conversation text with speaker labels

Recordings

Audio recordings for review and training

Summaries

AI-generated call summaries
Configure post-call extractions in the Extractions Tab to automatically extract:
  • Customer intent and sentiment
  • Key data points (name, email, order ID)
  • Custom fields you define

​
Agent configuration

Every Bolna agent is customized through 7 configuration tabs:

Agent tab

Prompts & PersonalityDefine your agent’s welcome message, instructions, and conversation behavior.

Intelligence tab

Intelligence & KnowledgeChoose your language model and connect knowledge bases for context-aware responses.

Languages tab

Voice & TranscriptionSelect voice provider, language, and transcription settings.

Engine tab

Latency, Behavior & HangupFine-tune response timing, interruption handling, user detection, the final call message, and call timeouts.
The former Inbound tab has been merged into Calling. Toggle on Inbound Calling there to pick an inbound number, match callers to a database and set spam limits — see Inbound Calling settings.

​
Next steps

What happens during a call

Follow one call from dial to webhook, with every status it passes through

Agent types

Choose between the agent types Bolna supports

Glossary

Look up any term used in the docs

Build your first agent

Generate a call-ready agent from a short description