Skip to content
Use case: incident response

When something breaks, an agent starts looking before you do.

A signed webhook from your monitoring wakes an agent. It reads the error in Sentry, checks the latest deployment on Vercel, and posts what it found to the on-call chat, with a suggested fix.

Approvals

Read-only tokens mean even an approved write cannot change anything.

Where it runs

Your own server, free, or Hosted on a private EU server from €29 a month.

Why CircleChat fits

  • Webhook triggers check an HMAC signature, and a JSON filter decides which events wake the agent.
  • The agent reads first and asks before any change, so it is safe at three in the morning.
  • Every run is recorded and can be replayed.

Set it up

  1. Step 1

    Connect the tools

    Add Sentry and Vercel or Cloudflare with read-only tokens where you can.

  2. Step 2

    Add a webhook trigger

    Create a trigger with a secret and a prompt template such as "New error: {{issue.title}}. Investigate."

  3. Step 3

    Point monitoring at it

    Send the alert to the trigger's address. The agent posts a trigger card and starts work.

Guardrails

  • Read-only tokens mean even an approved write cannot change anything.
  • DNS changes and texts ask first.
  • Requests without a valid signature are rejected.

Examples

  • Find the deployment that introduced a new error.
  • Text the on-call person through Twilio after approval.
  • Write the incident summary from the chat afterwards.
Questions

Short answers

What happens if a webhook has no secret?

Anyone who knows its address could call it, so always set a secret. Requests must then carry a valid HMAC-SHA256 signature.

Give your agents a seat at the table.

Your agents join the chat like any other teammate. Bring them in today.