Gemini CLI MCP: Connect Ken

In short Add https://mcp.ken.so/mcp as a remote MCP server in Gemini CLI, then sign in to Ken in your browser.

1. 01

   Add Ken as an HTTP Server

   Run this in your terminal. --scope user makes Ken available in every project; without it, Gemini CLI saves Ken to the current project only.

   gemini mcp add --scope user --transport http ken https://mcp.ken.so/mcp

2. 02

   Sign In with /mcp auth

   Start gemini and run /mcp auth ken. Your browser opens the Ken sign-in. Sign in with the account you use in the Ken app, then select Allow access on the Ken approval screen. You never paste a client ID or secret. Gemini CLI stores the token in ~/.gemini/mcp-oauth-tokens.json.

   /mcp auth ken

3. 03

   Check the Connection

   Run /mcp. Ken shows CONNECTED when the sign-in worked.

   /mcp

In short Yes. Gemini CLI can send your Ken API key as a Bearer token when a browser sign-in is not an option.

1. 01

   Use a Ken API Key Header Instead

   Create a key in the Ken app under Settings > Integrations > API Keys. Add Ken to ~/.gemini/settings.json with httpUrl, which is the Streamable HTTP key, and the key in headers. Use this on a server or over SSH, where Gemini CLI cannot open a browser.

   {
     "mcpServers": {
       "ken": {
         "httpUrl": "https://mcp.ken.so/mcp",
         "headers": {
           "Authorization": "Bearer <KEN_API_KEY>"
         }
       }
     }
   }

In short Ask in plain language. The client picks the Ken tool, and Ken runs it with your workspace role.

In short Most failures come from a skipped sign-in step or a client setting.