The vocabulary used across the platform UI, the docs and the API. When you see one of these words in your project, this is what it means.
Project & configuration
Tenant — your account. A billing and isolation boundary. One tenant can contain many projects, members and API keys.
Project — one business with one voice agent. Holds the scenarios, knowledge base, plugin connections, members, API keys and wallet draw. Most operations in the platform are scoped to one project.
Industry Pack — a calibrated bundle of everything an agent needs in a specific vertical (Restaurant, Sales, Real Estate). Picking a pack at project creation seeds the scenario, slots, intercepts, plugin requirements and sample knowledge. Fully editable afterwards.
Member — a person with access to a project. Roles: Owner, Editor, Viewer.
Role — the permission level a member has on a project. Owner = full control; Editor = configuration but no billing; Viewer = read-only.
Conversation runtime
Scenario — one call type for one project. "Inbound booking", "outbound reminder", "lead qualification" — each is a scenario. A project has many scenarios. Each is either inbound or outbound.
Session — one execution of a scenario for one phone call. Holds the call's transcript, slot values, tool calls, recording reference, outcome and cost.
Slot — a typed structured value the agent collects during a session. partySize: integer, bookingTime: datetime, phone: phone. Each scenario declares the slots it needs.
Goal — the business outcome a scenario is configured to achieve. Carries success criteria, constraints, escalation rules, tone, priority. Most scenarios have one primary goal.
Step — a named participant in a scenario's flow. Most simple scenarios are one slot_collection step that owns all the slots; the agent fills them in one flowing conversation.
Intercept — a named moment in the conversation where you can attach a custom handler. Silence, interruption, language switch, before-tool-call, slot-filled, before-session-end. Industry packs ship sensible defaults.
Outcome — what a completed session resolved to. Three values: success (goal achieved), middle (real conversation but no goal), spam (sub-30s drop or no engagement). Drives billing.
Commercial Intent Score — a 0-100 score the platform assigns each session after the call, indicating how commercially interested the caller seemed. Useful for prioritising follow-up.
Integrations
Plugin — an integration with an external system: telephony, booking, calendar, CRM, notification, storage. Plugins expose tools the agent calls during a session.
Plugin Type — the kind of integration (Telephony, Booking, Calendar, CRM, etc.). Every plugin of a type implements the same contract, so providers are swappable.
Tool Call — an invocation by the agent of a plugin function during a session. check_availability, create_booking, send_sms. Each tool call is recorded with its arguments, result and latency.
Knowledge base — the indexed content the agent uses to answer factual questions about your business. Sources: a website crawl, uploaded PDFs, direct-typed text.
Outbound
Outbound trigger — a request that initiates an outbound call. Sources: a UI click, a POST /v1/calls API request, a CSV batch upload.
Outbound batch — a CSV-uploaded set of contacts dialled at a controlled pace under one scenario. The batch page shows live progress and per-row outcomes.
Idempotency Key — your unique identifier for a state-changing API request. Repeating the same key returns the original response — safe for retries.
Do-not-call list — a per-project list of phone numbers blocked from outbound dispatch. Upload by CSV.
Webhooks & API
Webhook subscription — a URL we POST to when a specified event fires. Signed with HMAC, retried with exponential backoff.
Inbound webhook — a URL the agent exposes to receive backend events into active sessions. The payload is fanned out to every active session as an on_webhook_received intercept.
API key — a token your code uses to talk to the API. Project-scoped, permission-scoped, revocable. Shown once at creation; hashed at rest.
Scope — the operations an API key is allowed to perform. outbound:trigger, sessions:read, sessions:read:transcript, webhooks:manage, knowledge:manage.
Billing & usage
Wallet — your tenant's prepaid balance. Drawn against by call charges, crawler hours and chat tokens.
Top-up — moving money into your wallet, manually or by auto-top-up rule.
Tier — the provider stack a scenario runs on: Economy / Standard / Premium. Sets both the underlying models and the per-call rate.
Overage — minutes of a call beyond the 5-minute soft cap, when a scenario allows it. Charged per minute at the tier rate.
Sandbox — the project state where every charge is simulated. The wallet does not actually move; charges are clearly marked is_simulated. Default for new projects until you go live.
Usage event — a metered record (call minute, LLM tokens, STT seconds, TTS characters, telephony minute, storage GB-hour, API call, crawler second). Used to compute billing and surface per-project usage reports.
Mobile & operations
Daily digest — the morning summary the mobile app and email surface — what happened overnight and what needs your attention.
Take it / Let it run — the two actions on a live call from the mobile app: barge into the call as a human, or let the agent finish.
Goal-aware notification — a push notification that only fires for outcomes that need you (goal blown, escalation unhandled, plugin down, billing alert), not for every call.
Security & retention
Tenant isolation — the architectural rule that every data row carries a tenant id and every query enforces it. From outside it looks like each tenant has its own copy of the platform.
DSAR — Data Subject Access Request. A caller's right to know what data you hold about them and to demand deletion. The platform gives you the tooling to act within 30 days.
Audit log — the append-only record of privileged actions on the platform. Retained 7 years. Exportable.
Retention policy — how long the platform keeps your data. Recordings 90 days by default (configurable). Transcripts 12 months. Usage events 7 years. Audit logs 7 years.