Skip to main content
CodeRabbit’s knowledge base is the collection of context sources it draws on to make reviews smarter than a plain diff analysis. When CodeRabbit reviews a pull request, it combines several signals: what your team has taught it, the coding rules already in your repo, related code in linked repositories, external documentation fetched from the web, connected MCP servers, and issues or past PRs that provide business context. This section documents each context source, how to configure it, and when you need to take action to enable it.

​
Context sources

Learnings

Adaptive AI memory. CodeRabbit learns your team’s review preferences from natural-language chat and applies them automatically to future reviews.

Code Guidelines

Auto-detected team rules. CodeRabbit reads coding standards from .cursorrules, CLAUDE.md, .github/copilot-instructions.md, and other AI agent configuration files — no manual import required.

Multi-repo Analysis

Cross-repository context. Link related repositories so CodeRabbit can detect breaking changes, API mismatches, and dependency issues that cross repo boundaries.

MCP Servers

External tool context via the Model Context Protocol. Connect MCP servers to give CodeRabbit access to internal APIs, databases, or documentation systems during reviews.

Web Search

Up-to-date external documentation. CodeRabbit searches the web to gather context about libraries, APIs, and standards referenced in a pull request.

Linked Issues

Issue tracker context. CodeRabbit reads linked GitHub, GitLab, Jira, and Linear issues to understand the intent behind a pull request and validate that the changes address them.

​
What’s on by default vs. what needs setup

Context sourceStatusNotes
LearningsOn by defaultStarts learning from your first @coderabbitai chat interaction
Code guidelinesOn by defaultReads supported config files automatically; no setup required
Web searchOn by defaultEnabled unless explicitly disabled
Pull request contextOn by defaultUses past PRs for context; scope configurable
GitHub / GitLab issuesOn by default (private repos)auto scope; enabled for private repos, isolated for public
Multi-repo analysisNeeds setupRequires linking repositories in .coderabbit.yaml or the UI
MCP serversNeeds setupRequires an MCP integration to be configured
JiraNeeds setupRequires Jira integration and an active connection
LinearNeeds setupRequires Linear integration and an active connection

​
Issue trackers and past pull requests

CodeRabbit can pull context from issue trackers and your PR history using four knowledge_base configuration fields:
  • knowledge_base.issues — scope for GitHub and GitLab issues (local, global, or auto; default: auto)
  • knowledge_base.jira — use Jira as a knowledge source (auto, enabled, or disabled; default: auto); supports project_keys to restrict to specific projects
  • knowledge_base.linear — use Linear as a knowledge source (auto, enabled, or disabled; default: auto); supports team_keys to restrict to specific teams
  • knowledge_base.pull_requests — scope for past PR context (local, global, or auto; default: auto)
For Jira and Linear, auto means the integration is active for private repositories and disabled for public ones. See PR validation and linked issues for how CodeRabbit uses this context during reviews. See the configuration reference for all available options.

​
Global configuration

​
Opt out of data retention

All context sources that learn from your repository — including learnings, issues, and PR context — require CodeRabbit to retain data about your organization’s usage. To disable these features and remove any stored data, set knowledge_base.opt_out to true:
.coderabbit.yaml
# yaml-language-server: $schema=https://coderabbit.ai/integrations/schema.v2.json
knowledge_base:
  opt_out: true
Setting opt_out: true immediately and irrevocably removes all stored knowledge base data, including learnings. Stateless sources like code guidelines and web search continue to work.

​
Full configuration reference

For a complete list of knowledge_base options including scope values, file patterns, and integration keys, see the Knowledge base section of the configuration reference.