EmpathyC

Integrations & Alerts

From zero to monitoring in 10 minutes

A REST API, alert webhooks, and a three-step workflow designed for engineers who need to ship fast without cutting corners on safety.

01

Integrate API

12 lines of code

02

Configure Alerts

Email · Slack · Webhooks

03

Respond to Incidents

Your workflow, your control

Step 01

Integrate the API

Setup process

  1. 1Sign up at app.empathyc.co
  2. 2Choose tier — Starter / Growth / Scale
  3. 3Create integration (2 clicks)
  4. 4Download API key + endpoint URL
  5. 5Add 12 lines of code to your app

Language support

Any language with an HTTP client.

REST API — Python, JS, Ruby, Go, curl, whatever you ship with. Full docs →

Where to call it

Real-time

After every AI response — full continuous monitoring

Batch

Daily or hourly — historical analysis and trend reports

Async

Background job recommended — zero impact on user experience

Python example

analyze_conversation.py
import httpx

async def analyze_conversation(
    conversation_id: str,
    messages: list
):
    async with httpx.AsyncClient() as client:
        response = await client.post(
            f"{EMPATHYC_ENDPOINT}/v1/analyze",
            headers={
                "Authorization":
                    f"Bearer {EMPATHYC_API_KEY}"
            },
            json={
                "conversation_id": conversation_id,
                "messages": messages
            }
        )
        return response.json()
Step 02

Configure Alerts

Alert channels

EmailDefault — always on
SlackWebhook — recommended for teams
PagerDutyComing soon
Soon
TeamsComing soon
Soon

Alert triggers

Any safety flag

Immediate

Crisis, boundary violation, or harmful advice detected — immediate alert, no delay.

Crisis direct + indirect

Immediate

Both bypass org rate limits (20/hr for tier 1 flags). Crisis always gets through.

Every alert includes

EmpathyC Alert · Crisis Detected

  • →Conversation ID (opaque — your identifier)
  • →Crisis severity — none / indirect / direct
  • →Safety flags triggered with LLM reasoning
  • →3 quality scores — empathy, reliability, consistency
  • →Direct link to incident report in dashboard
Step 03

Respond to Alerts

When you get an alert

  1. 1Review the alert — email or Slack
  2. 2Open EmpathyC dashboard → full conversation + analysis
  3. 3Copy the conversation ID
  4. 4Look up the user in YOUR system — your database, your PII
  5. 5Verify the alert — real crisis or false positive?
  6. 6Take action per your SOP: call user directly, contact clinical staff, notify emergency services, document incident, or close as false positive

Intentional friction

Manual PII lookup forces human review. You retain control over data access and response decisions. In crisis scenarios, deliberate decision-making is the right default.

Responsibilities

Our responsibility

  • →Detect risks with high sensitivity
  • →Alert immediately with full context
  • →Provide evidence and reasoning
  • →Maintain zero-knowledge architecture

Your responsibility

  • →Verify alert accuracy
  • →Determine appropriate response
  • →Take action (or decide no action needed)
  • →Maintain duty of care per your jurisdiction

Technical FAQ

Common questions from engineers evaluating the integration.