AI voice agents that answer, decide and act — on every call.

API & Integrations

Everything the platform does on a call is available over REST, and every call event can be pushed to you by webhook.

Every call the platform handles is reachable over a REST API, and every call event can be pushed to your systems by webhook. Authentication is a bearer token issued from the portal; all responses are JSON.

Base URL https://api.kwinova.com/api/v1

What you can call

Area Endpoints What it does Status
Calls GET /calls · GET /calls/{id} · GET /calls/{id}/transcript List and filter calls, read a single call with its duration, outcome and per-turn latency, and pull the full transcript turn by turn. Available
Recordings GET /records The recording file for a call, with its timestamp and duration. Playback and download are also available in the portal. Available
Trunks & numbers GET /trunks · POST /trunks · PATCH /trunks/{id} Provision SIP trunks per realm, set priority and capacity, and manage the numbers attached to them. Available
Call flows GET /flows · PUT /flows/{id} Read and replace a flow document. Flows are validated on write, so a dead end or an unreachable node is refused rather than published. Available
Schedules GET /schedules · POST /schedules Time-of-day, weekday and holiday routing rules. Available
Knowledge search POST /search Hybrid full-text and vector search across your uploaded documents. Search-only API keys are available for read access from other systems. Available
Documents GET /documents · POST /documents · DELETE /documents/{id} Upload, list and remove the documents the agent answers from, and check processing status. Available
Providers GET /providers · GET /providers/{type} Which speech and language providers are configured, the languages and voices each offers, and their current health. Available
Health & metrics GET /health · GET /metrics Component health, active call counts and uptime. Metrics are exposed in Prometheus format. Available
Assistants POST /assistants · GET /assistants/{id} · PATCH · versions · rollback Named assistant objects — instructions, greeting, voice, outcomes and knowledge scope — with versioning and roll-back. In design
AI session control POST /calls/{id}/ai/start · /ai/cancel · GET /ai/status Start, cancel and inspect the AI session on a call in progress, including the current turn and live provider state. In design
Endpoints marked In design are specified but not yet released. Ask us before you build against them.

Calling your systems from a conversation

The API above is how your software talks to Kwinova. The more interesting direction is the other one: during a live call, the AI agent can call your endpoints as tools — look up an order, move an appointment, raise a ticket — and read the real answer back to the caller before the call ends. Each tool has its own timeout and fallback, so a slow API delays an answer rather than hanging the call.

Webhooks

Call started, call ended, outcome recorded, recording ready and transcript ready can all be delivered to an endpoint you own. Campaign calls can also be triggered in the other direction — post to the call-by-webhook endpoint and the platform dials, runs the flow and posts the outcome back.

Errors and limits

Errors return a JSON body with a stable code, a human-readable message and a details object. Standard HTTP status codes apply — 400 invalid request, 401 unauthorised, 403 forbidden, 404 not found, 429 rate limited, 503 unavailable.

Getting a token

Tokens are issued per company from the portal, with a read-only variant for search. Ask us and we will set up a sandbox realm with a test number so you can build against something real before you go live.