Bug: Magnific MCP OAuth flow fails — session state lost between authenticate and complete_authentication
Description
When using the Magnific MCP server (https://mcp.magnific.com/) via HTTP transport in Claude Code desktop app, the OAuth flow consistently fails with:
"No OAuth flow is in progress for magnific. Call mcp__magnific__authenticate first, then retry with the callback URL."
This happens even though authenticate was just called successfully moments before.
Steps to Reproduce
- Add the server:
claude mcp add --transport http magnific https://mcp.magnific.com/ - Call
mcp__magnific__authenticate→ receive auth URL - Open the URL in the browser, complete OAuth login
- Copy the callback URL (
http://localhost:<port>/callback?code=...&state=...) from the browser address bar - Call
mcp__magnific__complete_authenticationwith the callback URL
Expected Behavior
Authentication completes successfully and Magnific tools become available.
Actual Behavior
complete_authentication returns: "No OAuth flow is in progress for magnific."
Root Cause (suspected)
The MCP connection drops and reconnects between step 2 and step 5 (visible in the session as repeated "MCP server disconnected / reconnected" events). On reconnection, the OAuth session state is reset, so complete_authentication finds no active flow.
Tried 3+ times — same result every time. The issue appears to be a race condition between the MCP server's session lifecycle and the OAuth flow duration (the user needs a few seconds to log in and copy the callback URL).
Environment
- Claude Code v0.2.1
- macOS 15 (Mac Mini M4 Pro)
- Transport: HTTP (
--transport http) - MCP server:
https://mcp.magnific.com/
3 Comments
Found 1 possible duplicate issue:
This issue will be automatically closed as a duplicate in 3 days.
🤖 Generated with Claude Code
Closing this — turns out the issue was a misunderstanding on my end.
Magnific was already available as a
claude.ai Magnificconnector (authenticated via the claude.ai account, like Canva, Gmail, etc.). Theclaude mcp add --transport http magnific https://mcp.magnific.com/command added a redundant second server that was trying to do OAuth independently and failing.The OAuth failure is still technically real (the MCP session drops between
authenticateandcomplete_authentication, resetting the state), but since the integration already works through claude.ai connectors, it's not a blocking issue for us. Leaving this here in case it's useful for others who don't have the claude.ai connector and try to set up Magnific manually via CLI.This issue has been automatically locked since it was closed and has not had any activity for 7 days. If you're experiencing a similar issue, please file a new issue and reference this one if it's relevant.