October 2026¶
Explore the new resources available in October 2026. Learn how to expose Dataiku agents and tools through the native MCP server and build a conversational webapp with persistent history.
This page highlights selected major additions to the Developer Guide; routine updates, bug fixes, and minor improvements are not included. For product release notes, see our documentation.
Exposing tools and an agent through the native MCP server¶
Make your Dataiku agents and tools available to external applications through the native MCP server. In this new tutorial, you will build tools that retrieve information from Shortcut, create a visual agent to review an iteration’s remaining work, and configure MCP exposure at the instance and project levels. Use MCP Inspector to discover and call the exposed agent and tool.
Building a conversational webapp with an agent¶
The updated conversational webapp tutorial shows how to use the LLM Mesh conversational API to store conversation history in Dataiku. Connect a Python backend to a chat interface, retrieve each user’s messages, and let your agent answer follow-up questions after a page reload or backend restart. The tutorial includes complete Python, HTML, and JavaScript code, plus a way to archive a conversation and start a new one.
