2026-01-19 · Authensor

How to Get Email Alerts for AI Agent Safety Events

SafeClaw by Authensor sends email alerts for AI agent safety events, delivering detailed notifications about denied actions, policy violations, and audit integrity failures directly to your inbox. With support for SMTP, SendGrid, and Amazon SES, SafeClaw fits into any email infrastructure. It works with Claude and OpenAI agents, ships with 446 tests, and provides hash-chained audit logging.

Prerequisites

Step 1: Install SafeClaw

If you have not already, initialize SafeClaw in your project:

npx @authensor/safeclaw

Step 2: Configure SMTP Email

Add email settings to .safeclaw/policy.yaml:

version: 1
default: deny


notifications:
  email:
    provider: smtp
    host: "${SMTP_HOST}"
    port: 587
    secure: true
    auth:
      user: "${SMTP_USER}"
      pass: "${SMTP_PASS}"
    from: "safeclaw@yourcompany.com"
    to:
      - "security-team@yourcompany.com"
      - "devops@yourcompany.com"
    subject_prefix: "[SafeClaw]"
    events:
      - action.denied
      - audit.integrity_failure
      - policy.violation



rules:
  - action: file.read
    paths:
      - "src/**"
    decision: allow



- action: file.write
    paths:
      - "src/**"
    decision: prompt



- action: shell.execute
    decision: deny

Set environment variables:

export SMTP_HOST="smtp.yourcompany.com"
export SMTP_USER="safeclaw@yourcompany.com"
export SMTP_PASS="your_smtp_password"

Step 3: Use SendGrid Instead (Optional)

If you prefer SendGrid:

notifications:
  email:
    provider: sendgrid
    api_key: "${SENDGRID_API_KEY}"
    from: "safeclaw@yourcompany.com"
    to:
      - "security-team@yourcompany.com"
    template_id: "d-xxxxxxxxxxxxxxxxxxxxxxxxx"

SafeClaw passes structured data (action type, decision, target, agent, hash) as dynamic template variables, letting you design custom email templates in SendGrid's template editor.

Step 4: Configure HTML Email Content

SafeClaw generates HTML emails with structured information. Customize the template:

notifications:
  email:
    format: html
    template:
      header_color: "#FF4444"
      include_fields:
        - action_type
        - target_resource
        - decision
        - agent_identifier
        - timestamp
        - audit_hash
        - policy_rule
      include_audit_context: true
      context_entries: 5

The include_audit_context option includes the 5 most recent audit log entries surrounding the event, giving recipients immediate context for the alert.

Step 5: Set Up Daily Digest Reports

Instead of (or in addition to) per-event emails, configure a daily digest:

notifications:
  email:
    digest:
      enabled: true
      schedule: "0 8   *"
      to:
        - "engineering-leads@yourcompany.com"
      include:
        - total_actions_24h
        - denied_count
        - prompted_count
        - allowed_count
        - top_denied_actions
        - audit_chain_status
        - unique_agents

This sends a daily report at 8 AM summarizing all AI agent activity over the past 24 hours, including a breakdown of the most frequently denied actions and overall audit health.

Step 6: Test the Integration

Send a test email:

npx @authensor/safeclaw test-notify --channel email

Check your inbox for the test message. Then trigger a real denied action to confirm production delivery:

npx @authensor/safeclaw wrap -- node my-agent.js
npx @authensor/safeclaw audit --tail 5

The denial should trigger an email within seconds (or minutes, depending on your SMTP provider's delivery time).

Summary

SafeClaw integrates with any email infrastructure to deliver AI agent safety alerts. HTML-formatted messages with audit context give recipients immediate actionable information. Daily digests provide management-level visibility into agent behavior. SafeClaw is MIT licensed and open source.

Related Guides

Try SafeClaw

Action-level gating for AI agents. Set it up in your browser in 60 seconds.

$ npx @authensor/safeclaw