Hi jstaso,

We're not aware of any current issues with the MCP server. Which AI tool are you using with it? And are you seeing any error messages?

We'd love to dig into this.

I'm using Claude Sonnet 5. The last three days Claude says the connection is down.

I got this info from Claude:

Issue: Remember the Milk MCP connector failing after disconnect/reconnect
Details:

Disconnected and reconnected the RTM integration in Claude
After reconnecting, tool calls still fail immediately with a generic execution error (no specific error message returned — just "Error occurred during tool execution")
Tested with: rtm_list_tasks (a basic read call, limit 1)
Request ID from failed call: req_011CcfhEdkTznoqWWM9HsmxL
Timestamp: July 3, 2026
This is via Anthropic's Claude MCP connector for Remember the Milk

What I need: Confirmation on whether this is an auth/token issue on RTM's side post-reconnect, or an API/MCP server issue.

Hmm. I've been using the RTM MCP with Claude for the last couple days with no problem.

From Claude:

Just retried and it's still failing on my end — same generic error, new request ID this time: req_011CchjNEDpyT8gyCQ91ZeMs.
So the connector is definitely not working from where I'm sitting, even if RTM's status page or support team isn't seeing anything on their side. That's not unusual — this kind of failure could be in the MCP connector layer between Claude and RTM rather than RTM's core service itself, so it might not show up on their radar the same way.
A couple options:

If you're in touch with RTM support again, this second request ID might help them trace it further.

Hi jstaso,

We're looking into this. Would you happen to be using Claude on the web at claude.ai, or the Chat tab in the Claude desktop app?

We've been able to reproduce this using claude.ai and the desktop app, but strangely not Claude Code (or any other AI tools). We've also received a report from one other using who's using claude.ai.

We're looking into whether it's something that's happening on our end with claude.ai specifically, or if it's something on Anthropic's side.

I'm using on the web at Claude.ai.

I'm having this issue with the desktop app on Windows, android app, and claude.ai. I can confirm that it works fine from the Code CLI.

Just asked Claude what the issue could be... seems like there are multiple connectors that have this issue because code stores auth tokens locally, while the other methods of connection store the auth token on Anthropic servers associated with your account. Seems to be a broader Anthropic issue. Not sure what can be done to address this. Was really loving having Claude help me schedule my day based on RTM and my meeting calendar.

Same here, can't get the RTM MCP to work anymore in the Claude desktop app or the web app at Claude.ai. Reconnecting or removing and adding the MCP again does not help. I'm having no issues with the RTM MCP on ChatGPT.

The MCP is now working again with claude.ai on the web and Claude desktop.

The issue turned out to be on the claude.ai side. We've reported this Anthropic, but were able to implement a workaround on our end in the meantime.

If you experience any issues, let us know!