CodeRabbit Documentation - AI code reviews on pull requests, IDE, and CLI (original) (raw)

CodeRabbit Agent for Slack

Bring AI-powered investigation, planning, and action into your Slack workspace with shared controls, persistent knowledge, and governed access.

CodeRabbit Agent is an AI agent for your entire SDLC right in Slack. Investigate issues, generate implementation plans, discuss requirements with colleagues, pull in context from tools like Linear or Figma, and ask CodeRabbit to open a pull request — all from the channel or thread where the conversation is already happening, or in a direct conversation through the CodeRabbit Slack app. Context carries over between messages, durable facts are saved for future conversations, and every run can be reviewed after the fact.

What you can do

CodeRabbit Agent covers the full arc from investigation through implementation. Here are some of the things your team can do without leaving Slack.

Ask the Agent to explore your codebase, review merged PRs and design documents, and surface the approach your team already follows. Once the investigation is complete, the Agent can suggest changes and open a pull request, all within the same Slack thread.

Let the Agent pull telemetry from connected tools, correlate the data with recent code changes, and pinpoint the root cause. Once the team agrees on the fix, the Agent can open a surgical revert or patch PR.

When investigation turns into action, your team can continue the discussion in a thread — referencing context from connected tools like Jira, Linear, or Notion — and then ask CodeRabbit Agent to open a pull request that incorporates the full conversation, or create issues in your connected trackers.

Schedule tasks that CodeRabbit Agent can run on a cadence (run this task every day) or on each new matching channel message. Message-triggered automations can watch a channel for new alerts, notices, or other repeated inputs and reply in the triggering thread.

Core Concepts

The agent pulls your org’s context together, across every tool and team conversation, into one place.

• Code · The repositories, recent changes, and opened PRs.
• Tickets · Jira, Linear, or wherever your team tracks work.
• Docs · Notion pages, Confluence wikis, internal runbooks.
• Monitoring · Datadog, PostHog, Sentry, and your observability stack.
• Cloud · AWS, GCP, and the infra that runs under it all.

Set up external tools and APIs through Connections.

The agent builds a Knowledge base from Slack and your systems. Decisions, fixes, and patterns are captured as they happen and refined through daily use, so it reflects how your team actually works.

• Team knowledge · Org-wide know-how, shared across teams, channels, and repos.
• Channel memory · Retains team-specific patterns, runbooks, and conventions.
• Thread memory · Carries forward everything from the current task.

Learn more about how facts are stored and scoped in the CodeRabbit Agent Knowledge Base.

The agent works in a shared thread alongside your team. Anyone can guide, contribute, and move tasks forward. It learns from team conversations and stays aligned as work evolves.

• Synchronous · The agent lives in the Slack thread.
• Steerable · Redirect or expand the work mid-task.
• Resumable · Stop, switch devices, come back tomorrow.

See Working in Slack for how mentions, threads, and direct conversations work day to day.

Access, knowledge, and spend, scoped to the channel and user. Every run is explainable and attributed, so you see what the agent did, for whom, and what it cost.

• Access · Scoped to who you are and what you do. See Admin and security.
• Knowledge · Scoped to the channel where it belongs.
• Tool control · Decide which integrations are available where. See Scopes.
• Cost control · Attributed by team and channel, not individual keys. See Usage.

Billing

CodeRabbit Agent is billed separately from CodeRabbit review subscriptions. Your CodeRabbit plan covers pull request reviews, IDE and CLI reviews, and related features. CodeRabbit Agent usage is measured in agent minutes, the actual runtime each Slack Agent run spends working, and is billed independently. This means:

• A CodeRabbit Pro, Pro+, or Enterprise subscription does not include Agent usage.
• Agent charges are based on accumulated agent minutes, not on the number of seats or reviews in your CodeRabbit plan.
• You can use Agent without a paid CodeRabbit review plan, and vice versa.

Some Agent trials include a limited Agent Minutes grant. While that grant is active, CodeRabbit Agent uses the included minutes first. When the trial grant is exhausted or expires, CodeRabbit Agent pauses new work for the workspace until an admin activates the Agent plan. Track your agent-minute consumption from the Usage dashboard.

What’s next

Whether you are setting up CodeRabbit Agent for Slack for the first time or exploring what it can do, start at agent.coderabbit.ai and follow the pages below to get up and running.