2026-01-20 · Authensor

How to Send AI Agent Safety Alerts to Discord

SafeClaw by Authensor sends real-time AI agent safety alerts to your Discord server through webhook integration. When an AI agent action is denied, requires approval, or triggers an audit anomaly, SafeClaw notifies your team instantly with rich embedded messages. SafeClaw supports Claude and OpenAI agents, ships with 446 tests, and provides hash-chained audit logs.

Prerequisites

Step 1: Create a Discord Webhook

  1. Open your Discord server and navigate to the channel where you want alerts (e.g., #ai-safety-alerts).
  2. Click the gear icon to open Channel Settings.
  3. Go to Integrations > Webhooks.
  4. Click New Webhook, name it "SafeClaw Alerts", and optionally set an avatar.
  5. Click Copy Webhook URL. It looks like: https://discord.com/api/webhooks/1234567890/abcdefg...

Step 2: Configure SafeClaw for Discord

Add the Discord webhook to your .safeclaw/policy.yaml:

version: 1
default: deny


notifications:
  discord:
    webhook_url: "${DISCORD_WEBHOOK_URL}"
    username: "SafeClaw"
    events:
      - action.denied
      - action.prompted
      - audit.integrity_failure
    embed_color: "#FF4444"



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



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



- action: shell.execute
    decision: deny

Set the environment variable:

export DISCORD_WEBHOOK_URL="https://discord.com/api/webhooks/1234567890/abcdefg..."

Step 3: Customize Discord Embeds

SafeClaw sends Discord embed messages with structured fields. You can customize the appearance:

notifications:
  discord:
    webhook_url: "${DISCORD_WEBHOOK_URL}"
    username: "SafeClaw"
    avatar_url: "https://authensor.com/safeclaw-icon.png"
    embed_color_map:
      denied: "#FF4444"
      prompted: "#FFAA00"
      allowed: "#44FF44"
      integrity_failure: "#FF0000"
    footer: "SafeClaw by Authensor | Deny-by-Default"

Each embed includes the action type, target resource, policy decision, agent identifier, timestamp, and hash chain reference as separate fields.

Step 4: Route Events to Different Channels

For larger teams, send different event types to different channels:

notifications:
  discord:
    routes:
      - webhook_url: "${DISCORD_CRITICAL_WEBHOOK}"
        events:
          - action.denied
          - audit.integrity_failure
      - webhook_url: "${DISCORD_INFO_WEBHOOK}"
        events:
          - action.prompted
          - action.allowed

This sends critical events (denials and integrity failures) to a high-priority channel while routing informational events to a separate channel.

Step 5: Add Throttling

Prevent notification floods during heavy agent activity:

notifications:
  discord:
    webhook_url: "${DISCORD_WEBHOOK_URL}"
    throttle:
      max_per_minute: 15
      cooldown_seconds: 20
      batch_similar: true

The batch_similar option groups identical action denials into a single message with a count, reducing noise while preserving visibility.

Step 6: Test the Integration

Run the built-in notification test:

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

You should see a rich embed appear in your Discord channel with sample data. Then trigger a real denied action to confirm end-to-end delivery:

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

Check your Discord channel for the deny notification and verify the audit log:

npx @authensor/safeclaw audit --tail 5

Summary

SafeClaw delivers real-time AI agent safety alerts to Discord through webhooks with rich embedded messages. Event routing, color coding, and throttling give your team clear visibility without notification fatigue. Hash-chained audit logs back every notification with tamper-evident evidence. 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