Skip to main content
HunterAI ingests CRM data to power sales analytics and AI coaching. This guide is for a developer connecting a custom CRM — anything that is not amoCRM — to HunterAI. You implement a handful of read-only HTTP endpoints. HunterAI pulls from them on a schedule, maps the records into its canonical model, and everything downstream (dashboards, funnels, coaching) works without knowing which CRM produced the data.
This documentation is generated from and verified against the production adapter. Where an example shows a query parameter, a status code, or a field, that is what the adapter actually sends, reacts to, or reads.

​
What you build

Four required endpoints and up to three optional ones. All are GET, all return JSON in the same envelope, all are read-only.

GET /pipelines

Required · pipelines and their stages (with won/lost classification)

GET /users

Required · CRM managers

GET /contacts

Required · people

GET /leads

Required · deals

GET /events

Optional · lead history (stage moves, notes)

GET /tasks

Optional · manager tasks

GET /calls

Optional · call records

​
What you do not build

No webhooks

HunterAI pulls. You never push to us and never register a callback URL.

No write-back

The connection is one-directional. HunterAI never writes to your CRM.

No SDK

Plain JSON over HTTPS with a Bearer token. Any stack that serves HTTP works.

​
The contract in one screen

  • Transport — HTTPS GET against a base URL you host, e.g. https://crm.example.uz/api/hunterai/v1.
  • Auth — a long-lived read-only Bearer token you issue to HunterAI. See Authentication.
  • Envelope — every collection returns { "data": [ … ], "has_more": <bool> }. See How sync works.
  • Paging — HunterAI sends page and limit=100; has_more tells it when to stop.
  • Incremental — five endpoints accept updated_since; records come back ordered by updated_at ascending.
  • Errors — a small set of status codes drive retries and skips. See Errors.

​
Where to start

1

Read the Quickstart

Quickstart walks one endpoint end to end so you can see the exact shapes before building the rest.
2

Implement the four required endpoints

Use the Endpoint reference — each page lists every field the adapter reads, its type, and its maximum accepted length.
3

Check the Limitations

Limitations lists what HunterAI drops or collapses. Read it before you build — some constraints affect how you model currency and multi-contact deals.
4

Hand off to your HunterAI business owner

They create the connection and run the test call. See Go live.
Version 1.0. This spec is backed by a working adapter. Breaking changes will be recorded in the Changelog.