MCP integration

Connect Dewey to Claude, Cursor, and other MCP clients — hosted via OAuth or run locally over stdio.

Install hosted MCP#

Add the hosted MCP URL to Claude or Cursor and authenticate in-browser.

Hosted MCP is the easiest path — no local server, no API key in your client config. You add a URL to Claude Desktop, Cursor, or any MCP-compatible client and authorize the connection through a browser consent flow. The OAuth token is scoped per-org and revocable from the dashboard.

# Claude Desktop and Cursor both support custom MCP servers via URL.
# Add this URL in the client's MCP settings UI:
#
#   https://mcp.meetdewey.com/mcp
#
# The first request opens a consent screen at app.meetdewey.com.
# Approve, and the client receives a scoped OAuth token automatically.

Install local MCP#

Run the MCP server locally with a DEWEY_API_KEY.

Local MCP is useful if you want to avoid the hosted OAuth flow or run against a self-hosted Dewey. The server is published as `@meetdewey/mcp` on npm and runs over stdio. Drop the snippet into your client config file and your API key authorizes every call.

# Claude Desktop config locations:
#   macOS:   ~/Library/Application Support/Claude/claude_desktop_config.json
#   Windows: %APPDATA%\Claude\claude_desktop_config.json
#
# Add the dewey entry to mcpServers:

Response

{
  "mcpServers": {
    "dewey": {
      "command": "npx",
      "args": ["-y", "@meetdewey/mcp"],
      "env": {
        "DEWEY_API_KEY": "dwy_live_..."
      }
    }
  }
}

What the MCP server exposes#

The tools your agent gets when Dewey is connected.

Once Dewey is wired up, your MCP-compatible client gains a set of tools that map onto the REST API. Search and research are the most commonly invoked; the management tools come in handy when you want the agent to clean up its own corpus.