Now the agent moves first

It's 3:47am and Datadog alerts you of an issue with your checkout service. The #prod-alerts channel lights up and the engineer on call scrolls it, opens the runbook, pulls the dashboard, checks the last few deploys and starts typing the same kind of "here's what's happening" post they wrote last Tuesday.

And the Tuesday before.

Most of what happens after a Slack alert is muscle memory. You read the alert, decide if it's real, gather the context and then reply in thread with the next step.

So far, Automations in CodeRabbit Agent for Slack could run on a clock - every few minutes, hourly, daily, weekly. Perfect for "summarize last week's shipped changes every Monday." Not so perfect for the messy reality of incidents, deployments and customer pings, which don't run on a clock.

Something happens and then someone reacts. That's how most incidents work. So that's how the agent should work too.

Meet Triggers

Triggers are rules that start CodeRabbit Agent for Slack the moment a matching event lands.

Instead of waiting for a scheduled run, the agent fires when something real happens - a new alert in a channel, a Datadog event, a PagerDuty incident - and replies right in the triggering thread with whatever investigation, action, or next step you told it to take.

You create and manage them from the new Triggers tab inside the Automations page in the CodeRabbit Agent web app.

The four parts of a trigger

Every trigger is made of the same four pieces:

• Source: Where the event comes from (a Slack channel message, or a webhook from a service like Datadog)
• Matching rules: Which events should actually fire the trigger
• Agent instructions: What CodeRabbit Agent for Slack should do when it runs
• Destination: Where the agent posts the result

Today, you can set up two source types.

Slack channel messages

A Slack channel message trigger watches a specific channel for new top-level messages and fires whenever one matches the rule.

Use it when CodeRabbit should react to repeated operational events - new alerts, queue-health posts, deployment notices, customer escalations - where the same kind of message keeps showing up and a human keeps doing the same kind of follow-up.

You can match on:

• All new top-level messages in the channel
• Text filters using case-insensitive substring matching
• A required author allowlist for specific Slack bots or Slack apps

Two guardrails worth flagging:

• Only bots and apps can fire these triggers. Messages from human teammates never match, even if the text would otherwise qualify. The author allowlist is required, not optional.
• The agent always replies in the triggering thread. Same-thread delivery is enforced, so the investigation lives right under the alert that prompted it, not in some unrelated channel.

There's no Run now button for these. To test one, post a matching message in the channel from the allowlisted bot or app.

Example: React to new Datadog alerts

A reliability team wants CodeRabbit Agent for Slack to investigate each new Datadog alert posted in #prod-alerts. An engineer asks CodeRabbit to save a trigger that watches new channel messages from the Datadog bot, checks whether each alert looks real, and replies with the next step.

From that point on, every matching alert in #prod-alerts gets a thread reply from the agent. The trigger stays in that channel and only runs when a matching message arrives.

Webhook events

Triggers also accepts inbound webhooks for services that don't post to Slack as a bot. Supported providers include Datadog, PagerDuty, Pylon, and Custom webhook.

After saving a webhook trigger, CodeRabbit provides the webhook URL, required header, and a sample payload so you have everything needed to connect your source service in minutes.

You can match webhook events by:

• Provider event type
• JSON payload fields, or any combination of the two

Events that don't match any rule are dropped automatically, so nothing fires by accident.

Setting one up

1. Open Automations > Triggers in the CodeRabbit Agent for Slack web app
2. Pick the source - Slack channel message or one of the webhook providers
3. Define the matching rules
4. Write the agent instructions, exactly like what you'd otherwise type after @coderabbit in Slack
5. Pick the channel and delivery mode

The trigger goes live immediately. Like every other automation, it runs under the scope that applies to the channel where it lives, so it only ever reaches the repositories, connections, and spend limits that surface is allowed to touch.

Who can manage triggers

Workspace admins can view and manage all triggers, while other users can only see and manage their own. Message-triggered automations can't be run manually; to test one, post a matching message in the channel.

The Automations page also tracks run history, linking through to thread reviews so you can inspect exactly what the agent did, for whom, and how many agent minutes it used.

Try it out

Triggers are available now in the Automations page of your CodeRabbit Agent for Slack workspace.

New to CodeRabbit Agent? Head to CodeRabbit Agent to get started. New customers receive $50 per user in free agent minutes.

When that same alert wakes up your channel at 3 AM, save a trigger and let the agent take it from there.

Get started with CodeRabbit Agent for Slack