Add MCP servers in VS Code

Add MCP servers in VS Code

Add an MCP server when your agent needs a capability its existing tools don't provide, such as querying a database or retrieving information from an external service. Model Context Protocol (MCP) connects agents to these tools and data sources. If your existing tools meet your needs, you don't need another server.

This article takes you from adding a server to verifying its first tool call. Choose your starting point:

Before you start

  • Enable AI features and open a local project folder in VS Code.
  • Check the server publisher's prerequisites. A remote HTTP server can require an account or sign-in. A server launched by a command needs its runtime installed on the machine that runs it.
Caution

Local MCP servers can run code on your machine, and tools can access external services using your permissions. Review the publisher, command or endpoint, and requested access before installing or starting a server. Only trust configurations you have reviewed. Learn about MCP server trust.

Choose a server to try

For a concrete local example, use the Playwright MCP server. It requires Node.js 18 or newer, with npx available. Check your installation in a terminal:

node --version
npx --version

The walkthrough uses Playwright to demonstrate connecting an external tool provider. For everyday browser interaction, try the built-in browser tools first: they don't require an MCP server.

To try a remote documentation server instead, follow the Context7 setup requirements for its current authentication requirements and service limits. Use the HTTP endpoint https://mcp.context7.com/mcp in the connection steps below. A remote server doesn't require a local Node.js installation.

Connect a server

Use the publisher's URL or launch command. For the Playwright example, use the command in step two.

  1. Open the Command Palette (⇧⌘P (Windows, Linux Ctrl+Shift+P)) and run MCP: Add Server.

  2. Choose the option that matches what the publisher provides:

    For Playwright, choose Command (stdio) and enter:

    npx -y @playwright/mcp@latest

  3. Enter a server ID, such as playwright or context7, that isn't already used in the destination configuration. Use distinct IDs for different servers you plan to use together.

  4. For a personal configuration, select Global, and then Copilot Global when prompted. This stores the server outside your project. For team configurations or other harnesses, choose a configuration destination.

    If Copilot Global isn't available, check the Copilot harness requirements before continuing with this walkthrough.

  5. Review any server trust or authentication prompts before accepting them. Follow the publisher's sign-in instructions if the service requires authentication. Do not paste credentials into chat or commit them to your project.

You can also start the guided flow from Chat: Open Customizations > MCP Servers > Add Server. The editor is scoped to the selected agent harness.

For other ways to add a server, see Find another server.

Verify the first tool call

Adding a configuration isn't proof that the agent can use the server. Complete a request and inspect the tool call:

  1. In the agent session, enter a prompt that explicitly names the server and a low-impact action. For Playwright:

    Use the Playwright MCP server to open <your URL here> and report the page title. Use its tools, not the built-in browser tools. Do not change files.

    For Context7:

    Use the Context7 MCP tools to look up the React documentation for useState. Summarize the basic usage and cite the documentation returned by the tool. Do not change files.

  2. Review startup, sign-in, and tool approval requests as they appear. Approve only actions you understand.

  3. Expand the tool call in the conversation. Check that it belongs to the server you added and inspect its returned output.

Your connection works when all three checks pass:

  • A tool from the intended MCP server was called in your session.
  • The tool returned the requested page or documentation without a connection error.
  • The agent's answer reflects that tool output. A plausible answer alone isn't sufficient.

To control available tools in a Copilot session, open Chat: Open Customizations > Tools. For Local harness sessions, use Configure Tools in the chat input. See Manage available tools for the differences.

Find another server

Use the MCP server gallery (Preview) to browse publishers and capabilities:

  1. Open the Extensions view (⇧⌘X (Windows, Linux Ctrl+Shift+X)) and search for @mcp.

  2. Select a server and review its publisher, capabilities, and installation requirements.

  3. Select Install for a personal installation, or right-click and select Install in Workspace to share its configuration with the project.

  4. Complete the publisher's setup, and then verify a tool call.

If the gallery isn't available, use a publisher-provided URL or command with MCP: Add Server. Your organization can restrict which servers are available.

You can also open the gallery with MCP: Add Server > Browse MCP Servers.... Choose Add from another application... in the same picker to discover existing server configurations.

Add from a package or Docker image (model-assisted)

If you have a package or image name instead of a launch command, use model assistance to generate the server configuration:

  1. Run MCP: Add Server and select NPM Package, Pip Package, or Docker Image.
  2. Choose a configuration destination, then enter the package or image name when prompted. For example, enter @playwright/mcp for NPM Package.
  3. Review the package name, version, and publisher shown in the confirmation. Select Allow only if you trust the source, then follow the model-assisted setup prompts.
  4. Confirm a server ID that isn't already used in the destination configuration. Review the generated command, arguments, and credential requirements before accepting any server trust or tool approval prompts.
  5. Verify a tool call from the added server.

These options appear only when model-assisted setup is supported in your current configuration. If they aren't available, use the publisher's launch command with Command (stdio). Runtime prerequisites still apply: generating a configuration doesn't replace installing the required runtime.

Discover servers in the Customization Marketplace (Experimental)

For a Copilot session, turn on chat.customizations.marketplace.enabled Open in VS Code Open in VS Code Insiders , open Chat: Open Customizations, and select Discover. Select GitHub Feed from the source menu and search by name or with @type:mcp. Review the server and select Install.

Availability depends on the harness and your organization's marketplace policies. If automatic installation isn't supported, follow the publisher's instructions. Learn about marketplace discovery.

Choose a configuration destination

Choose a destination based on who needs the server and which harness runs it:

Your goal Configuration
Use a server across your workspaces Global > Copilot Global writes to $COPILOT_HOME/mcp-config.json, or ~/.copilot/mcp-config.json when COPILOT_HOME is unset.
Share a server with a team using compatible Copilot tools Workspace-root .mcp.json. The guided flow offers this destination when the Experimental chat.mcp.workspaceRootConfig.enabled Open in VS Code Open in VS Code Insiders setting is on.
Use a Local harness configuration with input variables or per-server sandboxing Workspace .vscode/mcp.json or user-profile mcp.json. These destinations remain available for compatibility and are labeled deprecated in the guided flow.

Portable files use a top-level mcpServers object. The VS Code format uses servers. Don't copy a configuration between them without checking the format and capability reference.

Important

Portable configurations don't support VS Code input variables or environment files. Servers requiring interactive inputs aren't forwarded to Copilot harness sessions. Use the authentication guidance for your configuration, rather than adding unsupported credential placeholders.

To reuse an existing setup from other applications, see automatic discovery. For Dev Containers or command-line installation, follow the environment-specific instructions.

Use context and prompts from your server

Servers can provide more than tools. Available capabilities depend on the server and your session:

  • Resources: attach server-provided data to a request from Add Context > MCP Resources, or run MCP: Browse Resources.

  • Prompts: use a server's prompt template by typing /<server>.<prompt> in chat.

  • MCP Apps (Experimental): supported servers can display interactive forms or visualizations in chat. See MCP Apps support.

If the server does not work

After correcting the problem, repeat the request and inspect the tool output again.

Manage an existing server

Continue with Manage and troubleshoot MCP servers to change a connection, stop or remove a server, review trust and startup behavior, or share configuration across environments.

For specific tasks, see MCP sandboxing, configuration synchronization, or enterprise access management.

Related resources

10/21/2026