MCP guide

Install an MCP server

Install Model Context Protocol servers from WorkFlowAI into Claude Desktop or Cursor using JSON config.

Browse ready configs on Skills & MCP. Each page includes install notes and a copyable configurationJson.

What is MCP?

Model Context Protocol (MCP) lets clients like Claude Desktop and Cursor talk to local or remote tools (filesystem, GitHub, databases) through a standard server interface.

1. Pick a server

  1. Open MCP servers on WorkFlowAI.
  2. Read the description, tags, and install command (npm / npx / binary).
  3. Prefer Verified entries when available.

2. Claude Desktop

  1. Open Claude Desktop → Settings → Developer → Edit Config.
  2. Merge the JSON from WorkFlowAI into your claude_desktop_config.json under mcpServers.
  3. Restart Claude Desktop.
{
  "mcpServers": {
    "example-server": {
      "command": "npx",
      "args": ["-y", "@example/mcp-server"]
    }
  }
}

3. Cursor

  1. Open Cursor Settings → MCP (or project .cursor/mcp.json depending on version).
  2. Add the same server block from the skill page.
  3. Reload the window if the server does not appear.

4. Environment & secrets

  • Put API keys in env vars referenced by the config — never commit secrets.
  • Filesystem / shell servers can access local files; scope paths carefully.
  • Review the GitHub repo linked on the skill page before production use.

5. Agent skills (not MCP)

Some catalog entries are agent skills (prompts and behaviors), not installable MCP servers. Use them as system prompts or skill files in your agent framework — they do not go in mcpServers.

Browse MCP servers MCP blog guide