MCP Directory
Back to Directory

MCP Servers for Cursor

Add MCP servers to Cursor via project or global mcp.json configuration.

Cursor IDE integrates Model Context Protocol servers so the AI agent in your editor can call external tools — read project files through a dedicated server, persist memory, fetch URLs, and more. Configuration patterns mirror other VS Code–family editors but Cursor exposes its own MCP settings UI and mcp.json file locations. PREREQUISITES Install runtimes required by the servers you add (from the official modelcontextprotocol/servers README): - Node.js 18+ for npx-based servers: @modelcontextprotocol/server-filesystem, server-memory, server-sequential-thinking, server-everything - uv / uvx for Python servers: mcp-server-git, mcp-server-fetch, mcp-server-time WHERE TO CONFIGURE MCP IN CURSOR Cursor documents MCP under Settings. The exact menu label may change between releases; look for "MCP" or "Model Context Protocol" in Cursor Settings. You typically have two scopes: 1. Global MCP config — applies to all projects (often ~/.cursor/mcp.json on macOS/Linux or equivalent in your user profile on Windows). 2. Project MCP config — .cursor/mcp.json in your repository root, shareable with teammates via git. Some Cursor versions also expose a graphical MCP panel to add servers without hand-editing JSON. If your build differs, prefer Cursor's in-app documentation — UI paths evolve faster than third-party guides. MCP.JSON FORMAT (COMMON PATTERN) Cursor follows the same command/args structure as Claude Desktop and VS Code MCP docs. A typical entry: { "mcpServers": { "filesystem": { "command": "npx", "args": [ "-y", "@modelcontextprotocol/server-filesystem", "${workspaceFolder}" ] } } } Note: VS Code MCP docs use a "servers" key in workspace mcp.json; Cursor commonly uses "mcpServers" in user-level config — verify against your Cursor version. When in doubt, add a server through Settings and inspect the generated JSON. WINDOWS — NPX WRAPPER On Windows, npx-based servers should use cmd /c per the official MCP servers README: { "mcpServers": { "filesystem": { "command": "cmd", "args": [ "/c", "npx", "-y", "@modelcontextprotocol/server-filesystem", "C:\\path\\to\\project" ] } } } Replace the path with your workspace directory. uvx commands do not need cmd wrapping. EXAMPLE: FILESYSTEM + MEMORY Verified install commands from https://github.com/modelcontextprotocol/servers : npx -y @modelcontextprotocol/server-filesystem /path/to/allowed/dir npx -y @modelcontextprotocol/server-memory Combined Cursor config (macOS/Linux paths): { "mcpServers": { "filesystem": { "command": "npx", "args": [ "-y", "@modelcontextprotocol/server-filesystem", "/Users/you/my-project" ] }, "memory": { "command": "npx", "args": ["-y", "@modelcontextprotocol/server-memory"] } } } After saving, restart Cursor or reload MCP servers from Settings so new processes spawn. CURSOR-SPECIFIC TIPS - Prefer project-scoped .cursor/mcp.json for repo-specific tools (e.g., Git server pointed at the current repository). - The Filesystem server supports MCP Roots on capable clients — directories can update at runtime without restarting the server. - Use Agent mode when you want the model to invoke MCP tools autonomously; verify tool permissions in Cursor's MCP settings. - Check Cursor changelog / docs if a config key is rejected — MCP support is actively developed. SECURITY Only allow directories and repositories you are willing to expose to automated tool calls. The filesystem server can write and overwrite files when tools like write_file are invoked. Reference servers carry MIT/Apache licenses but are not audited for production adversarial use. OFFICIAL REFERENCES - MCP reference servers and install commands: https://github.com/modelcontextprotocol/servers - Filesystem server tools and access control: https://github.com/modelcontextprotocol/servers/tree/main/src/filesystem - Cursor MCP documentation: use Help or docs.cursor.com for the latest UI steps (may vary by version).

Browse by category

File System

MCP servers that provide file system access for reading, writing, and managing files and directories. Enable AI assistants to interact with your local or remote file systems safely.

Databases

MCP servers for database operations including PostgreSQL, MySQL, SQLite, and MongoDB. Query, manage, and introspect databases directly from your AI assistant.

Web Search

MCP servers that provide web search capabilities. Enable AI assistants to search the web, retrieve real-time information, and access current data.

Developer Tools

MCP servers for developer productivity including code analysis, version control, CI/CD, and development environment integration.

Cloud Services

MCP servers for AWS, Azure, GCP, Cloudflare, Vercel and other cloud platforms.

Automation

Browser automation, scraping and workflow MCP servers for repetitive tasks.

AI & ML

Memory, reasoning, RAG, embeddings and machine-learning MCP servers.

Communication

Slack, Discord, email and messaging MCP servers for team communication.

Productivity

Notes, docs, calendars and task-management MCP servers.

Monitoring

Observability, logs, metrics and alerting MCP servers.

Security

Auth, secrets management and security tooling MCP servers.

Data Processing

ETL, analytics and data-pipeline MCP servers.