Content
summary Summary

French AI company Mistral is rolling out two major updates to its chatbot Le Chat: support for MCP-based integrations and a new memory feature that keeps track of conversation history across sessions.

Ad

With the new connector directory, Le Chat users can connect to more than 20 different platforms, spanning data analysis, software development, productivity, automation, and financial services. Integrations run on the Model Context Protocol (MCP), a standardized interface for large language models first introduced by Anthropic in late 2024 and now widely adopted as an industry standard.

According to Mistral, supported platforms include Databricks, Snowflake, Pinecone, Notion, Box, GitHub, Asana, Jira, Stripe, and Zapier. These connectors let users search and summarize data directly from Le Chat or trigger actions like creating Jira tickets and setting up tasks in Asana.

You can also use these connectors to build more complex workflows. For example, customer reviews could be analyzed in Databricks, tasks automatically generated from them in Asana, and results saved as summaries in Box. Or, you could check pull requests in GitHub, create related Jira issues, and document changes in Notion at the same time.

Ad
Ad

Companies can extend Le Chat with their own MCP connectors to support in-house tools. Role-based access control is supported, and self-hosting is an option.

Le Chat gets memory

With the new Memories feature, Le Chat can now remember information from past conversations. The goal is to deliver more relevant answers, personalized recommendations, and smoother interactions by recalling user preferences and decisions.

Mistral says only relevant information is saved permanently, while fleeting or sensitive data is ignored to avoid misinterpretation. The company is addressing common complaints about memory in systems like ChatGPT, especially concerns about unwanted context carryover in professional settings.

Whether this approach actually solves the problem remains to be seen. Mistral reports a retrieval accuracy of 86 percent, which sounds solid but becomes less impressive when factoring in the baseline errors of large language models.

Image: Mistral

Users can add, edit, or delete saved entries at any time. Importing memories from ChatGPT is also supported.

Recommendation

Both features—connectors and memory—are available to all Le Chat users, regardless of plan. Technical details and setup guides are available in the help section.

Ad
Ad
Join our community
Join the DECODER community on Discord, Reddit or Twitter - we can't wait to meet you.
Support our independent, free-access reporting. Any contribution helps and secures our future. Support now:
Bank transfer
Summary
  • Mistral is expanding its chatbot Le Chat with support for over 20 third-party tools, including Databricks, Snowflake, Notion, GitHub, and Asana, enabling direct integration and automation of complex workflows.
  • These integrations use the Model Context Protocol (MCP), which has become an industry standard and allows companies to connect their own tools through custom connectors, as well as use role-based access control and self-hosting.
  • The new "Memories" feature enables Le Chat to store relevant information from previous conversations to provide more targeted and personalized responses; users can manage these entries themselves and also import content from ChatGPT.
Sources
Matthias is the co-founder and publisher of THE DECODER, exploring how AI is fundamentally changing the relationship between humans and computers.
Join our community
Join the DECODER community on Discord, Reddit or Twitter - we can't wait to meet you.