Mistral AI is moving closer to introducing a memory system in its web application. A new banner has appeared in the interface, signaling that the feature is in preparation. According to the current design, memory will be accessible from the sidebar and marked with an alpha label, suggesting that it will launch in an early testing stage rather than as a finished product.
The feature is expected to function in line with other AI chatbots that already support memory. During conversations, key facts will be stored automatically, and users will have the ability to open a memory panel to review what has been saved. Importantly, users will also be able to delete or adjust saved memories, keeping control over what persists. This would allow more continuity between sessions, enabling the assistant to adapt to user preferences, context, or ongoing projects.
The banner appeared last week, which indicates internal rollout has started, though the timeline for public release remains uncertain. For Mistral, the move fits into a broader strategy of catching up with established players such as OpenAI and Anthropic, who already market memory as a differentiator in premium plans. If deployed, this feature could help Mistral make its platform more sticky for regular users, particularly those looking for personalization beyond single-session chats.
