Visual Studio Code’s Model Context Protocol (MCP) support has grown from agent access to external tools into a broader integration that can include prompts, data resources, authentication, and interactive components. Full MCP specification support was announced in June 2025, and GitHub said MCP support in VS Code 1.102 was generally available for production use with GitHub Copilot in July 2025. Whether you can use it in a particular organization still depends on administrator policy and, for preview features, the VS Code build you run.
What MCP support means in VS Code
MCP is a protocol layer that lets VS Code’s agent workflows communicate with external tools, services, and data sources through MCP servers. Instead of each client and service needing a separate integration, the protocol provides a shared way to expose capabilities to an MCP-compatible client.
For example, a server can give an agent access to browser automation, repository information, or API documentation. The agent can then use those capabilities as part of a task, subject to the server’s permissions and the client’s configuration. MCP does not make every server safe or automatically available: administrators and developers still need to choose and configure the integrations they allow.
What changed, and when MCP became available
- May 14, 2025: Microsoft described MCP as a consistent way for VS Code agents to reach external tools and services, citing Playwright MCP Server for frontend verification, GitHub MCP Server for repository insights and pull requests, and Context7 for API usage.
- June 12, 2025: VS Code announced support for the complete MCP specification, adding authorization, prompts, resources, and sampling to its existing support for tools and workspace awareness.
- July 14, 2025: GitHub announced that MCP support in VS Code 1.102 was generally available for production use with GitHub Copilot. Organizations must enable the MCP servers policy; it is disabled by default.
- September 2025, VS Code 1.105: VS Code documented an MCP marketplace in preview, powered by the GitHub MCP registry. The marketplace was disabled by default and controlled by the
chat.mcp.gallery.enabledsetting. - January 26, 2026: MCP maintainers announced MCP Apps as an official extension. Their client-support list identified Visual Studio Code support in Insiders, so MCP Apps should be treated as an Insiders or preview capability unless a later stable release confirms otherwise.
Which MCP features VS Code supports
The VS Code developer guide lists the following parts of the MCP experience. Availability of a server’s individual capabilities depends on what that server implements.
#1 Best Overall
- Tools: Agent mode can invoke tools exposed by an MCP server.
- Prompts: Servers can expose reusable prompts that appear in chat as slash commands in the form
/mcp.servername.promptname. - Resources and resource templates: Users can attach resources to chat or browse them. VS Code’s May 2025 release notes also describe saving resources from tool results, attaching them through Add Context, and browsing them with MCP commands.
- Elicitation: A server can request input from the user during an interaction.
- Sampling: A server can ask VS Code to request model work through the user’s configured models and subscription. The May 2025 release notes say first use requires confirmation and that users can inspect model controls and a request log.
- Authentication: The client supports authentication, including built-in GitHub and Microsoft Entra support.
- Roots and server instructions: MCP servers can receive roots and provide instructions to the client.
- Transports: VS Code supports local stdio, streamable HTTP, and legacy SSE transports.
- MCP Apps: These can return interactive UI components, but the cited January 2026 client-support information lists VS Code in Insiders.
How to add an MCP server safely
The exact setup depends on the server: a local server may use stdio, while a remote server may use HTTP and require authentication. The cited material establishes the supported transports and capabilities, but does not provide a universal server-add command or configuration recipe. Use the instructions for the specific server and the VS Code MCP developer guide rather than assuming every server has the same setup.
- Choose the server for the task. Check whether it provides the tools, prompts, or resources you need, and whether it is local, remote, or listed in the marketplace.
- Review its access and effects. Identify what data it can read, whether it can change external state, and what authentication it requires.
- Check organizational access. If you use GitHub Copilot under an organization or enterprise, ask the administrator whether the MCP servers policy is enabled. It is disabled by default.
- Configure and authenticate using that server’s instructions. Match the server’s requirements to a supported transport, and do not provide credentials to a server you have not vetted.
- Confirm the interaction at runtime. Pay particular attention to sampling requests, which require first-use confirmation, and to any tool that can alter a repository or another external service.
How to evaluate an MCP server
| What to check | Why it matters |
|---|---|
| Transport: stdio, streamable HTTP, or legacy SSE | It determines how the client communicates with the server and whether the integration is local or remote. |
| Authentication requirements | Confirm which account or credentials the server needs and whether your organization permits that access. |
| Read-only versus state-changing tools | Distinguish tools that inspect information from those that can create, update, or otherwise affect external systems. |
| Prompts and resources | Check whether the server exposes reusable prompts or data that can be attached to chat, not just callable tools. |
| Sampling behavior | Understand when the server may request model work through your configured models and subscription. |
| Data residency and privacy | Determine where data is processed and what information leaves your environment; do not infer these properties from MCP compatibility alone. |
| Administrator policy compatibility | Confirm that organization or enterprise policy permits the server and its intended use. |
| Distribution and maintenance | Note whether the server is local, remote, or discoverable through the preview marketplace, and assess who operates it. |
Marketplace and administrator controls
The MCP marketplace documented for VS Code 1.105 is a discovery feature, not a guarantee that a listed server is enabled or appropriate for a particular organization. In the cited release notes it was a preview, off by default, and controlled by chat.mcp.gallery.enabled. An organization’s MCP servers policy is a separate control: GitHub said it must be enabled for production use with Copilot and that it defaults to disabled. A user may therefore encounter a server or marketplace feature that is visible in documentation but unavailable under their current settings or policy.
Rank #2
What the expansion means for developers
The important shift is that MCP in VS Code is no longer just a way to call external tools. Its supported surface includes reusable prompts, attachable data, authentication, user-input requests, and server-initiated sampling, alongside multiple transports. That gives developers more ways to connect agent workflows to existing services, while also making permissions, data handling, and administrator governance central to deployment.
MCP co-creator David Soria Parra said: “I’m excited to see what developers will build now that they have access to the full spectrum of MCP features within VS Code. The community continually impresses me with their creativity and drive to push the boundaries of what’s possible.”
Free tools Windows power users keep installed
One-click scans. No signup required.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
