Skip to main content

Snowflake Cortex Code Remote MCP Connection Issues

When connecting Snowflake Cortex Code to the Atlan Remote MCP server, you may encounter issues related to server configuration, expired OAuth tokens, or the OAuth browser callback. This page covers the most common issues and how to fix them.

Atlan server doesn't appear in Cortex Code​

Error

No tools loaded from 'atlan' — the server is not listed among configured MCP servers.

Cause​

The server wasn't saved to the Cortex Code configuration file at ~/.snowflake/cortex/mcp.json, or it was added with an incorrect name, URL, or transport type.

Solution​

  1. Inspect the saved configuration:

    cortex mcp get atlan

  2. If the server is missing or incorrect, remove it and add it again:

    cortex mcp remove atlan
    cortex mcp add atlan https://mcp.atlan.com/mcp --transport http

  3. Restart the servers and confirm the connection:

    cortex mcp start

Atlan MCP stops working after OAuth token expires​

Error

MCP connection failed. The server is not responding — your OAuth token may have expired.

Cause​

OAuth tokens have a limited lifetime. Once the cached token under ~/.snowflake/cortex/mcp_oauth/ expires, the MCP server rejects requests and returns no results.

Solution​

  1. Remove the existing server:

    cortex mcp remove atlan

  2. Re-add the server:

    cortex mcp add atlan https://mcp.atlan.com/mcp --transport http

  3. Run cortex mcp start and complete the OAuth flow again in your browser.

Browser doesn't open or OAuth callback fails​

Error

OAuth authorization required for 'atlan' — waiting for authorization callback.

Cause​

The OAuth flow starts a local callback server (for example, on localhost:8585). If your browser doesn't open automatically, or a firewall or another process is using the callback port, the authorization never completes.

Solution​

  1. Copy the authorization URL from the terminal output and open it manually in your browser.
  2. Make sure no firewall or other process is blocking the local callback port.
  3. If the port is in use, stop the conflicting process, then run cortex mcp start again to restart the flow.

See also​

Need help​

If you need assistance after trying these steps, contact Atlan Support.