Skip to main content
| MCP (Model Context Protocol) servers are a knowledge source that CodeRabbit can query during reviews and chat. By connecting MCP servers to CodeRabbit, you give it access to context from your external tools — such as documentation systems, design files, and project management platforms — so that review comments and chat responses reflect your organization’s full context.
The number of MCP server connections depends on your plan: Essentials supports up to 5 MCP servers, Team supports up to 10, Advanced supports up to 15, and Enterprise supports up to 20.

​
What MCP servers add to reviews

When MCP servers are connected, CodeRabbit can draw on external context beyond the code itself. Supported integrations include:
  • Documentation and knowledge bases — Internal docs, wikis, and reference material
  • Project management tools — Issues, tickets, and project specs
  • Design tools — Figma designs and related design assets
  • Any tool with an MCP server — If a tool publishes an MCP server, CodeRabbit can connect to it without waiting for a formal integration
This makes review comments more relevant to your team’s standards and project context.

​
How CodeRabbit uses MCP during analysis

CodeRabbit acts as the MCP client — it ingests data from your connected MCP servers, not the other way around. During a code review or chat interaction, CodeRabbit:

Gathers context

Automatically calls relevant MCP tools to collect additional review context from your connected data sources

Enhances comments

Validates and enriches suggested review comments with the information retrieved

Supports chat

Accesses your organization’s knowledge bases and documentation when responding to PR chat questions

Stays transparent

Lists the tools it used under “Additional context used” in the review walkthrough
CodeRabbit searches your connected MCP tools automatically, but some tools organize their resources in ways that are hard to discover without a hint—for example, Jenkins Organization Folders or SonarQube projects with custom key formats. In those cases, add Usage guidance with URL patterns such as {repo} or {pr} to tell CodeRabbit where to look for each pull request. See User guidance for the full reference and examples.

​
Examples of MCP server types

Based on the integrations described in the CodeRabbit documentation, common categories of MCP servers you can connect include:
This approach unlocks integrations without waiting for formal CodeRabbit support. If a tool has an MCP server, CodeRabbit can connect to it.

​
Configuration

MCP servers must be added through the CodeRabbit web interface before they can be used. Open Connections, select Add connection, and choose a suggested MCP-compatible connector or the custom connector. For connection behavior and full setup instructions, see Connect MCP servers.
Once an MCP server is connected, control its usage with the knowledge_base.mcp section of your .coderabbit.yaml file:
.coderabbit.yaml
​
enum
default:"auto"
Use MCP servers as a knowledge source. ‘auto’ disables the integration for public repositories.
  • auto (default): Uses MCP servers as a knowledge source on private repositories and skips them on public repositories
  • enabled: Uses MCP servers on all repositories, including public ones
  • disabled: Never uses MCP servers as a knowledge source
​
array of string
Specify MCP server labels to disable (case-insensitive). These servers will be excluded from reviews and knowledge base queries.
Full reference: MCP For instructions on connecting and configuring MCP servers in the CodeRabbit app, see Integrate MCP servers.

​
What’s next

Set up MCP servers

Connect and configure MCP servers in the CodeRabbit app

Knowledge base overview

Learn about all knowledge sources available to CodeRabbit

Web search

Use web search as an additional knowledge source