Reliability data is only useful if it reaches the systems your team actually uses.

A call gets flagged for hallucination. Latency crosses your threshold on 10% of calls in the last hour. Tuner catches these in real time. Now it can tell the rest of your stack in real time too.

Voice AI teams don't all work in one place. Some route everything through a central alerting platform. Some have custom dashboards built exactly the way their team thinks. Others pull call data into a warehouse, a CRM, or a product they're building on top of voice.

The Tuner dashboard is a great place to understand what's happening in production. But often, you need those reliability signals to leave Tuner and land where your team already works.

That's exactly what Webhooks do.

Tuner now pushes real-time Voice AI events directly into your infrastructure, so your workflows can react the moment something happens.

Why we built this

Teams use Tuner in different ways. Some live in the dashboard. Others have a unified alerting stack they want everything to flow through, or they're pulling data out to build their own dashboards and workflows on top of Tuner. All of those are the point.

Tuner is a layer. You get the reliability and quality signal, and it's yours to surface wherever it's useful: route it, store it, alert on it, build with it. Webhooks is how that signal gets out of Tuner and into your stack.

The mechanics are simple. You register a URL. Tuner sends a POST when something happens. Everything downstream is up to you.

How webhooks work in Tuner

Setup takes less than a minute. Open your agent, navigate to Agent Settings → Webhooks, and paste the endpoint that should receive events. That could be your own API, an n8n workflow, a Zapier automation, or any endpoint capable of receiving HTTP requests.

Once saved, Tuner immediately starts sending events as they happen.

Every webhook follows the same structure regardless of event type, making it easy to build a single webhook handler.

{

 "event": "alert.fired",

 "event_id": "alert.fired:155:66",

 "created_at": "2026-07-06T03:10:35.868Z",

 "payload": {}

}

The event field tells you what happened. The event_id is unique per delivery, so you can deduplicate on retries. The payload carries the event-specific data.

Write one handler, switch on event, and you're done.

The events

1. call.analyzed

This event fires every time Tuner finishes analyzing a call, regardless of the voice platform.

Think of it as your "analysis complete" trigger.

The payload includes the call ID and a direct link back to Tuner. From there, you can retrieve the complete analysis through the API and build whatever workflow you need.

Common use cases include:

• Sync transcripts into your CRM

• Push call quality metrics into Snowflake or BigQuery

• Feed custom dashboards

• Trigger internal QA workflows

• Update customer records after every conversation
  
  

2. alert.fired

This is the event most production teams care about.

Whenever one of your Tuner alerts crosses its threshold, an event is sent immediately.

The payload includes everything your on-call system or an incident workflow needs to act immediately:

{

 "event": "alert.fired",

 "payload": {

   "alert_name": "hallucination-happened",

   "severity": "critical",

   "call_id": 8383,

   "agent_id": 466,

   "qualifying_count": 1,

   "threshold_n": 1,

   "window_ms": 3600000,

   "triggered_at": "2026-07-06T03:10:35.513Z"

 }

}

That means your incident platform immediately knows:

• which alert fired

• how severe it is

• which call caused it

• why it triggered

• when it happened

No polling. No refreshing dashboards. Just real-time production events.

3. call.received

For teams using Vapi or Retell, Tuner also exposes a call.received event.

This forwards the original provider payload unchanged, making it easy to plug Tuner into existing automations without changing your current workflows.

Alerts are more powerful than they look

alert.fired is only as good as the alerts you can define, and this is where Tuner goes deep. An alert in Tuner is not a simple "notify me." It is a live query over everything Tuner knows about your calls.

You can trigger on almost any field:

• Evals, like hallucination, intent missed, or identity not verified.

• Extracted data from the call itself, like whether data capture passed.

• Latency, at P50 or P90.

• Sentiment.

• Call metadata.

You can stack multiple conditions and combine them with ALL (AND) or ANY (OR). So a single alert can say: P50 latency is over 3000ms, and sentiment is negative, and data capture failed. Narrow and precise, or broad and catch-all, whatever fits.

Then you set how much is too much. Trigger on a raw count, like 10 or more matching calls, or on a percentage, like this happened in 10% of calls in the last hour. Over a window you choose, from 5 minutes to 24 hours. Add a severity, Info, Warning, or Critical, so downstream systems can route accordingly. More about alerts

That is a real monitoring engine. And now every one of those alerts can fire straight into your incident stack.

A real production example

Imagine you're operating a healthcare voice agent.

You've defined a critical alert that fires whenever hallucination is detected.

The moment a call qualifies:

• Tuner fires an alert.fired webhook

• PagerDuty opens an incident

• Slack receives the affected call

• A Jira ticket is created automatically

• Your engineering team can immediately investigate the exact conversation

Nobody has to refresh a dashboard or manually check production.

The system reacts automatically.

What you can build

Once events start flowing, the possibilities open up.

• Route critical production alerts into PagerDuty, Slack, or your on-call rotation.

• Build custom reliability dashboards in Grafana or your internal tools.

• Sync completed call analysis into Snowflake, BigQuery, or your data warehouse.

• Automatically create Jira tickets when production alerts fire.

• Push quality scores into Salesforce or HubSpot.

• Feed failed evaluations into your QA review pipeline.

• Trigger custom workflows inside your own Voice AI platform.

If you are building a platform on top of voice, or embedding voice AI into your product, this is the piece that lets Tuner become part of your infrastructure instead of a tab you check.

Get started

If you're new to Tuner, you can get started for free in just a few minutes. Sign up, connect your Voice AI agent, and Tuner will begin analyzing your production calls automatically.

Once your agent is connected, open Agent Settings → Webhooks, paste your endpoint, and enable the events you want to receive.

Pair Webhooks with one or two alerts, and you'll start receiving real-time Voice AI events within seconds.

For the complete event reference and setup guide, check out our docs.