belter.app
Get started
Platform · REST API

Connect any platform in days, not months.

If you don't see a native connector for your stack, push conversations in via REST. Single, bulk (up to 200), or CSV. Scoped tokens, idempotency, deterministic IDs. Genesys, NICE, Five9, Talkdesk, Webex, ServiceNow, Jira, and anything home-grown live here.

Generic Ingestion REST API CSAT Import
Get an API token Read the docs
curl · single conversation
POST /api/v1/ingest/conversations
Authorization: "Bearer btr_••••"
Idempotency-Key: "gen_847b21"

{
  "external_id": "genesys_a8d2",
  "channel": "chat",
  "brand": "aurora",
  "agent": { "email": "[email protected]" },
  "customer": { "id": "c_4127" },
  "messages": [
    { "role": "customer", "text": "Can't access cPanel" },
    { "role": "agent",    "text": "On it — give me 2 mins." }
  ]
}

// → 202 Accepted · evaluation queued
Endpoints

Three shapes. One conversation model.

Push one, push up to 200, or paste in a CSV — Belter normalizes everything into the same conversation + message + customer schema before evaluation.

Verb Endpoint Purpose Limit
POST /api/v1/ingest/conversations Single conversation ingest. Idempotent by external_id or Idempotency-Key. 1
POST /api/v1/ingest/conversations/bulk Bulk array ingest. Each item evaluated independently — partial success returned. 200
POST /api/v1/ingest/conversations/csv CSV upload for historical backfills. Streamed, no chunk limit. —
POST /api/v1/ingest/amazon-connect Dedicated ingest for Amazon Connect Contact Lens / CTR payloads. 1
GET /api/v1/conversations/{id} Fetch a conversation incl. score, criteria reasoning, sentiment timeline. —
GET /api/v1/evaluations List evaluations with filters: agent, brand, score range, date. —
POST /api/v1/csat Single CSAT score. Correlated to conversation by external_id. 1
POST /api/v1/csat/bulk Bulk CSAT — drain a Qualtrics, SurveyMonkey, or Typeform export. 500
POST /api/v1/csat/webhook Endpoint your survey tool calls. Auth via header token. —
Authentication

Scoped tokens. Per-org. Per-purpose.

Every token is bound to one org and one or more scopes. conversations.write for ingest, evaluations.read for reporting, etc. Mint a fresh token for each integration, rotate at will, revoke without affecting the others.

  • Bearer auth · Authorization: Bearer btr_…
  • Scopes · least privilege. We list the scopes a token has at the bottom of every error response.
  • Idempotency · Idempotency-Key header dedups retries. external_id deduplicates conversations.
  • Rate limits · 1,000 RPM per token, 200 conversations per bulk call. We send X-RateLimit-* headers.
  • Encrypted at rest · payloads + recordings + tokens all encrypted; tokens hashed with HMAC-SHA256 before storage.
Available scopes
# conversation pipeline
conversations.write     # push new
conversations.read      # fetch + list
evaluations.read        # scores + reasoning
evaluations.trigger     # force re-evaluate

# auxiliary signals
csat.write              # /api/v1/csat
agents.read             # roster
alerts.read             # smart alerts

# administration
webhooks.manage         # subscribe events
org.read                # settings
CSAT Import API

Pull satisfaction scores into the same picture.

Most QA platforms treat CSAT as an island. Belter ties survey scores back to the underlying conversation so you can compare what your customers thought with what your QA rubric thought.

POST /api/v1/csat — single
{
  "external_id": "genesys_a8d2",
  "score": 4,
  "scale": "1-5",
  "verbatim": "Sara was super patient",
  "received_at": "2026-06-08T14:32:00Z"
}

// → 200 OK · linked to conv_84291
CSAT linked to QA score ● Auto-correlate
CSAT 5 "Super helpful, sorted my migration" 94
CSAT 4 "Took a while but got there" 88
CSAT 2 "Felt rushed, didn't really answer" 68
Belter flags QA-CSAT divergence — high QA + low CSAT = something the rubric is missing.
Read API

Pull scores back out for your BI tools.

Conversations, evaluations, agents, CSAT, alerts, webhooks — all listable and filterable via scoped read tokens. Plug into Looker, Mode, dbt, or whatever lives upstream of your dashboards.

GET /api/v1/evaluations
GET /api/v1/evaluations
   ?brand=aurora
   &score_lt=70
   &from=2026-06-01
   &limit=100

// → 200 OK · 42 evaluations
{
  "data": [
    {
      "id": "eval_a8d2",
      "conversation_id": "conv_84291",
      "score": 62,
      "rubric_version": 7
    }
  ],
  "next": "cursor_..."
}
GET /api/v1/conversations/{id}
// includes:
{
  "id": "conv_84291",
  "channel": "chat",
  "agent": { "name": "Jordan T." },
  "messages": [ ... ],
  "evaluation": {
    "score": 94,
    "criteria": [
      { "name": "empathy", "score": 97 },
      { "name": "resolution", "score": 92 }
    ],
    "sentiment_timeline": [ ... ],
    "pii_flags": []
  }
}
Already supported

If it has an export, we can ingest it.

Our integration engineers have wired up dozens of platforms via the REST API. Most go from kickoff to first scored conversation in under a week.

Genesys Cloud CX NICE CXone Five9 Talkdesk Webex Contact Center ServiceNow ITSM Jira Service Management Twilio Flex Help Scout Front Kustomer Gladly Dixa Sprinklr Verint Custom / homegrown
Keep exploring

Other integration categories

OutboundWebhooksStream events from Belter to you. Live ChatLiveChat, ChatwootNative chat connectors. SupportZendesk, Intercom, FreshdeskNative ticket connectors. VoiceRingCentral, Amazon ConnectPhone call ingest.
Build with us

Stand up an integration this afternoon.

Get a sandbox token, push a single conversation, watch it score. The whole loop, in less than the time it takes to write a JIRA ticket about it.

Get an API token Talk to engineering