---
title: Glossary
description: "Short definitions of every term the Callab AI dashboard uses, from Agent Identity to Workspace, in alphabetical order."
---

**Agent.** An AI that holds a conversation on your behalf, by phone (voice agent) or
by text (text agent). See [Core concepts](/get-started/core-concepts).

**Agent Identity.** The first prompt field: who the agent is, its tone and role.

**Advisor.** The page, under **Observability** in the sidebar, that lists
configuration issues detected across your agents, knowledge bases, campaigns and
numbers.

**AMD.** Answering-machine detection. Notices when a voicemail picks up an outbound
call and reacts instead of talking to a recording.

**Auto-recharge.** A billing setting that buys minutes automatically when the balance
falls under a threshold.

**Background audio.** An ambient sound played under the agent's voice, such as a
call-centre room tone, so the call sounds like a place rather than a vacuum.

**Balance.** The minutes and text messages left in the company account, shown at the
bottom of the sidebar.

**BYOT.** Bring your own Twilio. Adds a phone number from your own Twilio account.

**Call.** One conversation with a voice agent, from the browser or the phone, with
its recording, transcript, summary and outcomes.

**Call quality.** A score on each call's page, alongside AI relevancy and
faithfulness, judged after the call from the transcript.

**Campaign.** A calling job: outbound (the agent calls a contact list) or inbound (the
agent answers a number).

**Category.** A free-text grouping on a contact, such as Lead or Customer. Campaigns
can target it.

**Change request.** A prompt change drafted by AI from thumb-down feedback, shown
under **Change Requests** to review, edit and accept. Accepting creates a new
revision.

**Company.** Your account. Holds members, workspaces, billing, phone numbers and
storage.

**Connection.** A CRM connected through OAuth under **Integrations**, such as HubSpot
or Pipedrive.

**Contact.** A person with a phone number. The audience of outbound campaigns.

**Conversation.** One text-agent thread: its turns, tokens and tool calls.

**Denoiser.** Noise reduction applied to the caller's audio before the agent hears
it.

**Deployment.** A text agent published to Slack, Telegram, Microsoft Teams, WhatsApp
or Discord.

**Dynamic variable.** A placeholder in the prompt, such as `{{customer_name}}`,
filled per conversation.

**Evaluation.** A test scenario with steps and success criteria, run against an agent,
with a pass rate per run.

**First message.** What the agent says when it speaks first.

**Flow.** A prompt type where the conversation is a graph of nodes rather than one
prompt.

**Guardrails.** The third prompt field: style rules, phrases to use or avoid,
boundaries.

**Inbound.** A call the agent receives. Also a campaign type.

**Integration.** An automation flow under **Integrations**: inbound flows pull data
in, outbound webhooks push data out.

**Knowledge base.** Documents the agent can search during a conversation through
the Knowledge Base tool.

**Launch Checklist.** The pre-ship checks in the agent workspace, shown in
**Publish**. A required check that fails blocks publishing.

**MCP.** Model Context Protocol. A tool type that connects an agent to an MCP server's
tools and data.

**Modality.** Voice or text. Fixed when an agent is created.

**Outbound.** A call the agent places. Also a campaign type.

**Outcome.** A structured field extracted from each conversation once it ends.

**PII redaction.** Removes names, emails, phone numbers and other personal details
from transcripts, summaries and analytics before they are stored.

**Press Keys (DTMF).** A tool that presses keypad digits during a call, to get through
an IVR menu or confirm a prompt.

**Prompt type.** Single prompt or flow.

**Publish.** Takes the agent's saved version live, once the required launch checks
pass. Saving alone does not.

**Recommendation.** A next step for an agent, ranked by expected impact, under
**Recommendations**. Re-evaluated after every revision.

**Revision.** The second number in a version, `v0.3`. Increments on every save and
starts again when the agent is published.

**Schedule Callback.** A tool that books a callback at a time the caller chooses,
placed later by an outbound campaign.

**SIP trunk.** A carrier connection that brings your own phone numbers to Callab AI.

**Tag.** A label on a contact. Campaigns target tags.

**Tool.** An action the agent can take during a conversation: end or transfer the
call, hand it to another agent, press keys, schedule a callback, call a webhook, send
an email, book on Cal.com, push to Zapier, search a knowledge base or use an MCP
server.

**Turn detection.** How the agent decides the caller has finished speaking and it is
its turn.

**VAD.** Voice activity detection. The thresholds that decide what counts as speech
and what counts as silence.

**Version.** A saved state of an agent, numbered and diffable, with rollback.

**Widget.** The chat or call interface embedded on a web page.

**Workspace.** A partition of a company that holds agents, campaigns and contacts.