[BUG] Todoist MCP loses session with "No transport found for sessionId" error

Preflight Checklist

• [x] I have searched existing issues and this hasn't been reported yet
• [x] This is a single bug report (please file separate reports for different bugs)
• [x] I am using the latest version of Claude Code

What's Wrong?

Todoist MCP connection drops mid-session after 20-30 minutes of use. When attempting to use Todoist tools, Claude Code returns "No transport found for sessionId" error. Interestingly, HA-MCP (another HTTP-based MCP server) remains stable in the same session, suggesting per-server transport state loss rather than global timeout. The /mcp reconnect option doesn't resolve the issue - only a full Claude Code restart restores functionality.

What Should Happen?

MCP server connections should remain stable throughout the session, or automatically reconnect when transport errors occur. If reconnection fails, the /mcp menu reconnect option should successfully restore the connection without requiring a full application restart.

Error Messages/Logs

Error: Streamable HTTP error: Error POSTing to endpoint: {"jsonrpc":"2.0","error":{"code":-32001,"message":"No transport found for sessionId"},"id":null}

The Todoist MCP connection is having issues.

Steps to Reproduce

1. Configure Todoist MCP: claude mcp add --transport http todoist https://ai.todoist.net/mcp
2. Authenticate via /mcp menu
3. Use Todoist MCP tools successfully (e.g., create tasks, add comments)
4. Continue working for 20-30 minutes, using multiple MCP servers (both Todoist and another HTTP-based MCP)
5. Attempt another Todoist tool call
6. Observe "No transport found for sessionId" error
7. Note that HA-MCP remains functional in same session
8. Attempt /mcp reconnect - doesn't resolve issue
9. Restart Claude Code - Todoist MCP works again

Claude Model

Opus

Is this a regression?

I don't know

Last Working Version

NA

Claude Code Version

2.1.19

Platform

Other

Operating System

macOS

Terminal/Shell

Terminal.app (macOS)

Additional Information

• Regular Claude web interface (claude.ai) uses Todoist MCP successfully with no transport issues
• Another HTTP-based MCP server (private HA instance) remains connected while Todoist drops
• This suggests HTTP/SSE transport state is managed per-server and some servers lose session context
• Pattern occurs consistently after 20-30 minutes of mixed MCP usage
• Related to known SSE timeout issues (#3033, #10129) but distinct error signature