Glossary
A comprehensive reference of terminology used in Neuratel Studio, the API, and voice AI in general.
A
Agent
An AI-powered entity configured to handle voice phone calls. Each agent has a brain (LLM), voice (TTS), transcriber (STT), conversation settings, tools, and optional analytics.
API Key
A secret token used to authenticate requests to the Neuratel API. All keys are prefixed with nk_live_.
B
Background Audio
Ambient sound played during a call to make the conversation feel more natural. Examples: office noise, coffee shop ambiance.
Backchannel
Short acknowledgment sounds an agent makes while the caller is speaking (e.g., "mm-hmm", "I see"). Signals active listening without interrupting.
Barge
A supervisor action where you join an active call and both the caller and agent can hear you. See also: Listen, Whisper.
Brain
The LLM configuration of an agent. Includes provider (OpenAI, xAI, Groq, Phantom), model, temperature, max tokens, and system prompt. This is where the agent's personality and instructions live.
BYOC (Bring Your Own Carrier)
Connect your existing telephony carrier to Neuratel via SIP trunking instead of using Neuratel-provisioned numbers.
C
Call List
A collection of contacts to be dialed in an outbound campaign. Created by uploading a CSV file.
Campaign
An automated outbound calling initiative. Links an agent to a call list and dials contacts sequentially with configurable concurrency and retry logic.
Contact
An individual entry in a call list. Contains a phone number and optional metadata (name, company, etc.) used as template variables.
Conversation
The configuration block controlling how an agent behaves during a call — first message, silence timeout, max duration, interruption handling, and turn detection settings.
I
Interruption
When a caller speaks while the agent is still speaking. Configurable: agents can stop immediately, finish the current sentence, or ignore the interruption.
K
Knowledge Base
A collection of documents (PDF, DOCX, TXT, or URLs) that an agent can search during a conversation using RAG (Retrieval-Augmented Generation).
L
Listen
A supervisor action where you hear both the caller and agent but neither can hear you. See also: Whisper, Barge.
M
MCP (Model Context Protocol)
A standard protocol for connecting external tools to AI agents. Agents discover and call tools from an MCP server during live conversations — e.g., looking up a customer record or booking an appointment.
Member
A user belonging to an organization. Each member has a role (Admin, Operator, Viewer) that determines their permissions.
N
Node
A single step in a workflow. Nodes represent actions like "call agent", "send SMS", "wait", or "branch on condition".
O
Organization
The top-level workspace in Neuratel. Contains agents, phone numbers, team members, credentials, and billing.
P
Post-Call Analysis
Automated evaluation that runs after a call ends. Includes sentiment analysis, success evaluation, and rubric scoring powered by an LLM.
R
RAG (Retrieval-Augmented Generation)
A technique where the agent searches a knowledge base for relevant context before generating a response. Improves accuracy for domain-specific questions.
Role
A permission level assigned to a team member. Options: Owner, Admin, Operator, Viewer.
Rubric
A set of scored criteria used in post-call analysis to evaluate whether the agent achieved its objectives. Each criterion has a name, description, and pass/fail threshold.
S
Scope
A permission granted to an API key. Scopes control which endpoints and operations a key can access. Examples: agents:view, agents:call, analytics:view, campaigns:create.
Sentiment Analysis
Automated detection of the caller's emotional tone (positive, negative, neutral) during and after a call.
SIP (Session Initiation Protocol)
The standard protocol for initiating and managing voice calls over the internet. Used for BYOC (Bring Your Own Carrier) integrations.
Success Evaluation
A post-call analysis check that determines whether the agent achieved its goal (e.g., "Did the agent successfully book an appointment?").
T
Template Variable
A placeholder in an agent's system prompt that gets replaced with contact-specific data during outbound campaigns. Syntax: {{variable_name}}. Example: {{contact_name}}.
Tool Calling
The mechanism by which an agent invokes external functions (via MCP) during a conversation. The LLM decides when to call a tool based on the conversation context.
Transcriber
The STT (Speech-to-Text) configuration of an agent. Includes provider (Deepgram), model, and language. Converts the caller's speech into text for the LLM.
Transfer
Moving a call from an AI agent to another destination (human operator, external number, or another agent).
Turn Detection
The mechanism that determines when a speaker has finished talking. Uses VAD (Voice Activity Detection) and silence thresholds to detect turn boundaries.
V
VAD (Voice Activity Detection)
Technology that detects when someone is speaking versus when there is silence. Used for turn detection and interruption handling.
Voice
The TTS (Text-to-Speech) configuration of an agent. Includes provider (ElevenLabs, Cartesia, Phantom Speech), voice model, stability, and speed settings.
W
Warm Transfer
A transfer where the agent introduces the caller to the human operator before disconnecting. The agent stays on the line during the handoff.
Blind Transfer
A transfer where the agent connects the caller directly to the destination without an introduction. The agent disconnects immediately.
Webhook
An HTTP callback that Neuratel sends to your server when events occur (call ended, recording ready, analysis complete).
Webhook Signing Secret
A cryptographic key used to verify that webhook requests originated from Neuratel. Used with HMAC-SHA256 signature verification.
Whisper
A supervisor action where you speak only to the agent during a live call — the caller cannot hear you. Used to coach the agent in real time. See also: Listen, Barge.
Workflow
An automated sequence of actions triggered by events or schedules. Built using the visual workflow builder with nodes and connections.
Transfer Types Comparison
| Type | Agent Stays? | Introduction? | Use Case |
|---|---|---|---|
| Blind Transfer | No | No | Simple routing to a known destination |
| Warm Transfer | Yes, until handoff | Yes | Complex situations requiring context |
Acronym Quick Reference
| Acronym | Full Name |
|---|---|
| BYOC | Bring Your Own Carrier |
| BYOK | Bring Your Own Key |
| LLM | Large Language Model |
| MCP | Model Context Protocol |
| RAG | Retrieval-Augmented Generation |
| SIP | Session Initiation Protocol |
| STT | Speech-to-Text |
| TTS | Text-to-Speech |
| VAD | Voice Activity Detection |
Next Steps
- FAQ — common questions answered
- Concepts — deeper explanations of core concepts
- Troubleshooting — solutions to common issues