Skip to content
Pro

Model Context Protocol

This page only applies to Raycast for Mac V1. Latest information for Raycast for Mac V2, Windows, and iOS is available in the New Raycast Manual.

Model Context Protocol or MCP is an open protocol that standardizes how applications provide context to LLMs. Raycast is a client that can connect to servers. Servers can be local data sources such as the file system or remote services such as external APIs.

Screenshot of the Playwright Browser Automation MCP in Raycast V1

You can use MCP servers within Raycast to extend AI even further. MCP servers work and behave similar to our AI Extensions. After installation, you can @-mention MCP servers in the root search, AI Commands, AI Chat, and Presets.

Install MCP Servers

MCP Servers are decentralized — shared on GitHub and other places. You can simply install a server with the Install Server command by filling out the form.

Screenshot of the Install MCP Server command in Raycast V1

For now, we only support the stdio transport.

When running servers we pass the PATH from the default SHELL to the process. If you can run the command (npx etc) in your default shell in a terminal, then it should run in Raycast also. If you edit your PATH then you need to re-start Raycast to pick up the change.

If you copy a MCP configuration JSON before opening the Install Server command, Raycast attempts to input the form. Try it with the following configuration:

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

Deeplink

You can also programmatically install a MCP Server with the following deeplink:

raycast://mcp/install?<configuration-json-percent-encoded>

The JSON configuration has to be in the following format:

FieldDescriptionExample
nameName of the server"Sequantial Thinking"
typeServer connection type"stdio"
commandCommand to start the server executable. The command needs to be available on your system path or contain its full path."npx"
argsArray of arguments passed to the command.["-y", "@modelcontextprotocol/server-sequential-thinking"]
envEnvironment variables for the server.{"API_KEY": "…"}

Manage MCP Servers

After you installed a few MCP Servers you can search them with the Manage Servers command. The command allows to uninstall servers or start a new chat with them.

Screenshot of the Manage Servers command in Raycast V1

Use MCP Servers

MCP Servers are used in the same way as our native AI Extensions. Simply @-mention a server after installation to use it with AI.

Screenshot of using MCP servers in Raycast V1

Explore MCP Servers

To discover servers, we built a meta registry. It contains official servers that are maintained by companies, community ones that are showcasing the power of deeply integrated services into AI, as well as an integration with Smithery.

Screenshot of Search Servers command in Raycast V1

The registry also works with other clients such as Claude or Cursor. This makes it very easy to explore and install MCP servers across the ecosystem.

FAQ