Nurtra API v1

Authenticate every request with an organisation API key, minted under Settings → API & Integrations in the CRM:
  Authorization: Bearer nk_live_...

Or sign in as a person (the mobile app's way) and use the session token the same way:
  POST /v1/auth/login    {"email": "...", "password": "...", "device": "iPhone 15 · Nurtra 1.0"} → {"token": "nk_user_...", ...}
  POST /v1/auth/login/google     {"idToken": "...", "device": "..."} → the same, from a Google sign-in on the phone
  POST /v1/auth/login/microsoft  {"idToken": "...", "device": "..."} → the same, from a Microsoft sign-in on the phone
  POST /v1/auth/login/apple      {"idToken": "...", "authorizationCode": "...", "nonce": "...", "device": "..."} → the same,
                                 or 401 apple_not_linked with a linkToken to send with one of the sign-ins above
  POST /v1/auth/logout   ends that session; a password reset or a disabled account ends it too

Endpoints:
  GET  /v1/me            who this key or person is — organisation, permissions, centres, connected Apple ID
  POST|DELETE /v1/me/identities/apple   connect or disconnect the signed-in person's Apple ID
  GET  /v1/centres       the centres the key may see
  GET  /v1/pipeline      the pipeline's stages
  GET  /v1/leads         families, newest first (search, stage, centreId, createdSince, page, pageSize) — search is the
                         CRM's search box: family, guardian or child name, phone or email, best match first
  GET  /v1/leads/{id}      one family
  POST /v1/leads         create an enquiry (Idempotency-Key header honoured; duplicates answer 409)
  GET  /v1/tours         every family's tour, soonest first (status, centreId, from, to, order, page, pageSize)
  GET  /v1/tours/{leadId}  one family's tour — the same shape, with the centre, host, notes and outcome
  GET  /v1/waitlist      the children waiting for a place, longest wait first (centreId, order, page, pageSize)

Website chats with prospective parents, for signed-in people who work families (a key has nobody to answer as):
  GET  /v1/chats                 conversations, newest activity first (status=open|closed|all, waiting=true, centreId, page, pageSize)
  GET  /v1/chats/{id}            one conversation with its messages — after=<messageId> for only the newer ones, which is how to poll
  POST /v1/chats/{id}/reply      {"body": "..."} → the reply, under the person's name; the family's widget shows it within seconds
  POST /v1/chats/{id}/close      end the conversation; a family on record gets the transcript on their timeline
  POST /v1/chats/{id}/lead       {"leadId": "..."} → tie the conversation to a family on record
  A family's message queues a push to the registered phones (data.type "chat", with the threadId) — once per stretch of waiting.

The dashboard's AI, for signed-in people (a key has nobody to answer for):
  GET  /v1/ai/ask               the ask bar's suggested questions, and whether a model is switched on
  POST /v1/ai/ask               {"question": "..."} → an answer from the person's own pipeline, with pages worth opening
  GET  /v1/ai/findings          the findings card — kind=dashboard (default) | consultants | demand
  POST /v1/ai/findings/refresh  regenerate the findings now, as the card's Refresh button does

Rate limit: 120 requests a minute per key, over a one-minute window. Past it the
API answers 429 with an error whose code is "rate_limited" and a Retry-After header saying how many
seconds to wait. Page through /v1/leads and /v1/tours with page and pageSize (up to 200) rather than
polling one family at a time. The AI endpoints are limited separately, to 20 a minute
per session, because each may reach the model.

Try it in the browser: https://api.nurtra.ai/docs
Reference (OpenAPI):   https://api.nurtra.ai/openapi/v1.json