Skip to content
Harmful deploymentSerious1 sourceFirst seen
Picture from the github.com article
Image: github.com

claude-mem uploads your last 30 days of prompts to a server you never agreed to

Ourtake

Unticking the checkbox did not stop the upload, and the provider that received the prompts cannot be identified.

What happened

The claude-mem installer's setup screen read local chat histories and uploaded the last 30 days of users' prompts to cmem.ai by default, with no opt-in and no working way to stop it. Prompts were routed through OpenRouter to a DeepSeek model, and cmem.ai says it cannot determine which provider received them.

Source