Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.nixflex.com/llms.txt

Use this file to discover all available pages before exploring further.

After every call, Nixflex runs an analysis pass over the transcript. You get a summary, the caller’s sentiment, whether the call succeeded, and any custom data you’ve defined — automatically. No setup required for the defaults. Everything is generated by a second Claude call after the conversation ends.

​
What you get out of the box

Every call record has these fields populated automatically:
FieldTypeDescription
call_summarystring2–4 sentence summary of what happened
caller_sentimentenumhappy, neutral, or frustrated
call_successfulbooleanDid the agent achieve its goal?
extracted_dataobjectCustom fields you configured (see below)
Example response from GET /v1/calls/:id: 
{
  "call_id": "CA1234567890",
  "call_summary": "Caller booked a check-up appointment for Tuesday 8th March at 2pm. Confirmed contact details and texted address.",
  "caller_sentiment": "happy",
  "call_successful": true,
  "extracted_data": {
    "caller_name": "Sarah Johnson",
    "email": "sarah@example.com",
    "appointment_date": "2026-03-08",
    "appointment_time": "14:00"
  }
}

​
Custom data extraction

Beyond the defaults, you define fields you want pulled from each call. Add an extraction_schema to the agent: 
{
  "extraction_schema": {
    "caller_name": "The caller's full name if mentioned",
    "email": "Email address if provided",
    "appointment_date": "Booked date in YYYY-MM-DD format",
    "appointment_time": "Booked time in 24-hour HH:MM format",
    "service_requested": "Type of service (check-up, cleaning, whitening, etc.)",
    "callback_requested": "true if the caller asked us to call them back"
  }
}
For each call, Claude reads the transcript and fills in whatever it can. Missing fields come back as null — Claude won’t invent data.

​
When analysis runs

  • Inbound and outbound calls — both directions, always
  • All calls over 5 seconds — calls that ended in voicemail or instant hangup are skipped to save cost
  • Asynchronously, after the call ends — usually within 5–10 seconds
By the time your webhook fires, all analysis fields are populated.

​
Querying for outcomes

The dashboard’s Call History page is one way. The API lets you filter calls by analysis results: 
# All successful calls today
curl "https://api.nixflex.com/v1/calls?call_successful=true&since=2026-05-17" \
  -H "Authorization: Bearer KEY_ID:KEY_SECRET"

# All frustrated callers
curl "https://api.nixflex.com/v1/calls?caller_sentiment=frustrated" \
  -H "Authorization: Bearer KEY_ID:KEY_SECRET"
Useful for:
  • Quality assurance — review frustrated calls
  • Lead routing — flag successful sales calls for follow-up
  • Reporting — aggregate sentiment over time
  • Data exports — populate your CRM with extracted fields

​
Cost

Post-call analysis adds roughly $0.001 to the cost of each call (one Claude Haiku call over the transcript). It’s included in the per-minute pricing — you don’t pay extra.

​
Disabling extraction

To skip extraction for an agent (rare; e.g. very high call volume where you only need the call recording): 
curl -X PUT https://api.nixflex.com/v1/agents/agent_xxx \
  -d '{"enable_post_call_analysis": false}'
The transcript and recording are still saved; just the AI fields are skipped.

​
Reading analysis programmatically

Each call object includes the analysis inline. There’s no separate endpoint to fetch it. See Get call for the full schema.