How to Set Up MCP in Cursor

MCPModel Context Protocol. A standard that lets an AI agent pull in context from outside the repo, like Jira tickets or internal docs. Press Enter for the full definition. is an open standard for connecting AI agents to external tools and data sources. In Cursor, an MCP server exposes tools that the agent can call in the middle of a task, such as querying a database, reading a ticket or calling an API. With them, the agent can work with more than your codebase.

A Cursor growth engineer describes MCPModel Context Protocol. A standard that lets an AI agent pull in context from outside the repo, like Jira tickets or internal docs. Press Enter for the full definition. servers as a great way to connect Cursor to internal knowledge bases, and sorts them into two buckets. Personal servers are tools only you use, such as Figma or a meeting-notes service like Granola. Team servers are the shared systems your organization reads from: a data warehouse such as Databricks, an experiment platform such as Statsig, or your backend services. Once a server is connected, you can steer a confused agent mid-run by naming it: "use the Databricks MCP."

The two buckets leave out a cost on the agent's side. Every enabled tool is one more option the model weighs on each turn. Cursor's docs suggest switching off the servers you don't need to cut that clutter. Keep the personal bucket small, and switch team servers on for the work that uses them instead of leaving all of them on.

Point Cursor at the server, give it any keys it needs, and confirm its tools show up. A local server gets a command to launch, and a hosted one gets a URL.

1. 1For a one-click install from the Cursor Marketplace, open Customize in the sidebar, choose MCPs and click Add to Cursor on the server. For anything else, create a .cursor/mcp.json in your project.
2. 2For a manual entry, add a command (stdio, for example npx -y some-mcp-server) or a remote URL (streamable HTTP or SSE).
3. 3Provide any required environment variables or API keys. Marketplace remote servers can authenticate with OAuth instead.
4. 4If you edited the config file, save it and restart Cursor. Then check that the server's tools appear in the tools list at the top of the chat panel.
5. 5Ask the agent to use a tool. By default, Cursor asks for your approval first. MCPModel Context Protocol. A standard that lets an AI agent pull in context from outside the repo, like Jira tickets or internal docs. Press Enter for the full definition. tools follow the same run mode as terminal commands, so an allowlisted tool can run without asking.

{
  "mcpServers": {
    "example": {
      "command": "npx",
      "args": ["-y", "@example/mcp-server"],
      "env": { "API_KEY": "..." }
    }
  }
}

Decide which file the entry belongs in before you write it. A global ~/.cursor/mcp.json follows you into every project, and the project file wins when both define the same server name. So a global entry can sit there shadowed while you edit it and wonder why nothing changes.

The project file also lives inside the repository. An API key pasted into env is one commit away from going out with every clone. Write ${env:API_KEY} in place of the key, and Cursor reads the value from your environment. Marketplace remote servers authenticate over OAuth, which keeps the key out of the file.

The servers worth connecting are the systems your team already works in. The table lists common ones and what the agent can do with each.

To see what exists, browse the Cursor Marketplace for official plugins first, then cursor.directory for community servers. Between them, they show whether a given tool, such as Excalidraw or Lucidchart, has a server yet.

The Figma MCPModel Context Protocol. A standard that lets an AI agent pull in context from outside the repo, like Jira tickets or internal docs. Press Enter for the full definition. works in both directions, which is why designers reach for it. Inbound, you paste a Figma file link into the prompt with a request such as "use the design system at this link to inform the design of my dashboard." Cursor reads the tokens, typography, colors and variable definitions, takes screenshots, and applies them to your live code.

Outbound, you select a rendered component in Design ModeA way to point at an element in Cursor's built-in browser and change it directly, instead of describing it in words. Press Enter for the full definition., press ⌘+L to send it to chat, and say "publish this component back to Figma" with the target file link. Cursor recreates the component there with the same colors, fonts and width, split into proper layers. The published layers show up in purple.

Almost always, it means the server never started. The usual causes are a bad command or path, a missing API key, or a mismatch between stdio and remote. Run the exact command in a terminal and watch it start before you blame Cursor.

• The command or path is wrong, or the package isn't installed. Test the command in a terminal first.
• Environment variables or API keys are missing, so the server fails to start.
• A remote server was given a command. Remote servers need a URL.
• If the tools load but the agent keeps picking the wrong ones, too many may be enabled across servers. Trim the list to what you need.
• Still stuck? Open the Output panel (⌘/Ctrl+⇧+U) and choose MCPModel Context Protocol. A standard that lets an AI agent pull in context from outside the repo, like Jira tickets or internal docs. Press Enter for the full definition. Logs. The server's startup errors land there.

Start with the terminal test, because it splits the problem in two. If the command fails in a terminal, the fault is in the command, the package or the keys, and Cursor's config is not involved. If it starts in a terminal but not in Cursor, compare the command and environment variables in your config with what you ran.