Skip to main content
A lot of engineering work starts the same way every time. A new alert needs a first look, a new ticket needs someone to find the likely cause, and a channel needs its weekly summary of what changed. Each time, someone has to notice the event, dig through the code and the other tools, and write up what they found. An automation hands that job to CodeRabbit Agent, the AI agent that investigates your code and connected tools, answers questions, and opens pull requests. You describe the job once, and CodeRabbit runs it every time its trigger fires, with no one asking. Use an automation when the same request would otherwise come up again and again, for example to:
  • Assess each new Datadog or PagerDuty alert against recent pull requests and deploys, and suggest the next step.
  • Investigate each new Linear, Jira, or GitHub issue and comment with the likely cause and a suggested fix.
  • Implement a fix and open a pull request when an issue gets a label.
  • Post a weekly summary of customer-visible changes to a Slack channel.
A run starts on a schedule, on a new message in a Slack or Discord channel, or on a webhook event from Datadog, Sentry, PagerDuty, Google Cloud, Linear, Jira, GitHub, GitLab, Intercom, Zendesk, or Pylon, or from any service that can send a webhook. The result goes wherever the automation can write: a Slack or Discord channel, a pull request in a GitHub repository, or any connected tool, such as a comment on the Linear issue, a note on the PagerDuty incident, a page in Notion, a record in HubSpot or Salesforce, or an MCP server. Each run uses the repositories and connections that the automation’s scope allows, and its result is listed on the automation’s page. No Slack or Discord app is required: a Linear issue can be the trigger and a GitHub pull request the result.
The page of an example automation named Carrot Patch Patrol, created by @thumper: when a new issue lands in the Burrow team in Linear, dig through the code and comment with the likely cause and a fix. Stat tiles show 10 runs, a 100% success rate, an average duration of 149.8 seconds, and a last trigger 1 day ago. The configuration shows the trigger, on matching Linear webhook events, and the action, delivers to 1 destination. Recent runs list comments on issues BUN-69, BUN-68, and BUN-67, such as the carrot counter going negative when two bunnies harvest at once.

An example automation: a Linear webhook trigger, one destination, and what each run posted

How an automation works

Every automation has three parts, which match the three cards in the automation editor: When the trigger fires, CodeRabbit starts a run, follows the instructions, and delivers the result. You can follow each run on the automation’s page and in Automations > Activity. See Manage and monitor automations.

Where you can create automations

All surfaces save to the same list. An automation created in Slack or Discord shows up in the web app, and you can edit it there, except one-time automations.

Examples

A platform team wants a daily report of new dependency vulnerabilities on the main branch. An engineer creates an automation in the web app with a Scheduled trigger that runs every weekday at 9:00 in the team’s timezone, and these agent instructions:
The destination is the #platform-eng Slack channel. After saving the automation as active, the engineer selects Run now to check the output right away. From then on, a fresh thread with the results appears in the channel each weekday morning.
For complete setups of common automations, such as turning labeled Linear issues into pull requests, incident response, and ticket triage, see Automation guides.

Who can use Automations

Automations aren’t part of a CodeRabbit plan. An Essentials, Team, Advanced, or Enterprise subscription neither includes nor blocks them: runs are billed separately as Agent minutes, like the rest of CodeRabbit Agent. The Automations Agent usage card at the top of the Automations page shows whether automations without a Slack or Discord destination can run. When they can’t, it says why, such as billing not set up, an expired trial, or a reached spend cap, and Go to billing opens Billing > Plans & products. Until then, these automations can’t be created or resumed, and their runs are recorded as failed. To cap spend for one team or channel, see Spend limits.

Automations and similar features

Other CodeRabbit features also run work automatically. They are configured separately:

What’s next

Create and edit automations

Build your first automation in the web app, from chat, or with the API.

Automation triggers

Choose between schedules, channel messages, and webhook events from 12 providers.

Automation guides

Set up incident response, support triage, ticket triage, and more, step by step.