Recommended MCP servers

Warning

This feature is only available in Visual Studio Code.

Even though Code Assistant can connect to any Model Context Protocol (MCP) server that matches the specification, the community has already put together several high-quality out-of-the-box servers. This page lists the recommended servers and step-by-step guides on how to configure them for you to quickly get going.

MCP servers

Name	Description
Context7	Documentation for the agent on the majority of popular libraries for development with code examples. With Context7, the agent writes better code.
SourceCraft MCP	MCP server with a toolkit for intelligent use of the SourceCraft platform. Comes with tools you need to work with pull requests, issues, milestones, etc.
Yandex Cloud Data Catalog Consumer	Tool for the user of the centralized metadata repository of an organization. Allows searching for different types of metadata (tables, views, and queries) and build lineage graphs at the table and column level.
Yandex Cloud Documentation	MCP server for real-time access to official Yandex Cloud documentation via generative search.
Yandex Cloud Toolkit	Toolkit used to deploy simple apps in Yandex Cloud. Includes Yandex Compute Cloud, Yandex Virtual Private Cloud, Yandex Identity and Access Management, Yandex Object Storage, and Yandex Managed Service for YDB.
Yandex Search	Web search tool based on Yandex Search API, both generative and classic. The use of search is billed based on the Yandex AI Studio pricing policy.

Tip

For up-to-date information on Yandex Cloud MCP servers, refer to the Yandex Cloud MCP Servers repository.

Installing MCP servers

Installation from Marketplace

In the chat's top panel, click Marketplace and go to the MCP tab.
Select the MCP server of interest and click Install.

Warning

Consider the prerequisites and fulfill them before installing the MCP server.
Select server Installation Scope:
- Global: The MCP server will be available in all workspaces.
- Project: The MCP server will be available at the git repository level.

Select the installation method and install the server.

Manual installation

Let's try installing an MCP server manually with Context7 as an example.

Global configuration

In the chat's top panel, click MCP Servers.
Click Edit Global MCP.
In the JSON file that opens, paste the following configuration to the mcpServers object and save the changes:

macOS/Linux

Windows

{
  "mcpServers": {
    "context7": {
      "command": "npx",
      "args": ["-y", "@upstash/context7-mcp@latest"]
    }
  }
}

{
  "mcpServers": {
    "context7": {
      "type": "stdio",
      "command": "cmd",
      "args": ["/c", "npx", "-y", "@upstash/context7-mcp@latest"]
    }
  }
}

In Windows, you may need to call npx for the cmd.exe shell.

Project configuration

If you prefer to add your configuration to a repository, create a file named .codeassistant/mcp.json in the project root and add the same fragment:

macOS/Linux

Windows

{
  "mcpServers": {
    "context7": {
      "command": "npx",
      "args": ["-y", "@upstash/context7-mcp@latest"]
    }
  }
}

{
  "mcpServers": {
    "context7": {
      "type": "stdio",
      "command": "cmd",
      "args": ["/c", "npx", "-y", "@upstash/context7-mcp@latest"]
    }
  }
}

Note

If both the global and project files define the server with the same name, the project configuration has a higher priority.

Checking the installation

Make sure the Enable MCP Servers option is enabled in the MCP settings panel.

Context7 should appear in the drop-down list of servers.
To start the server, if it is not started yet, click .
Code Assistant will request your permission when first calling Context7. Approve the permission to continue.

What's next

Expand the drop-down list next to the MCP server and check the list of tools provided with Context7.
Configure auto-approving actions for tools that you use most often.

Was the article helpful?

What makes MCP different from the REST API

Chat