TimeXtender Data Integration Winter 2026 Release

TimeXtender has released a new version of their products with a winter theme: major Snowflake improvements. On top of that, there's also a big leap forward with the TimeXtender MCP Server and Deliver endpoints. In this blog I'll highlight seven of the key updates; for the full details of the TDI and connector releases, you can check the following links:

• TDI 7256.1 Release details
• Data Source Providers release

Snowflake

Until now, when using Snowflake as the engine for the Prepare instance, it was necessary to set up Ingest storage via ADLS. As of this release, it's now also possible to use Snowflake native staging and store raw data in Snowflake. This results in less infrastructural complexity, especially for environments with restrictive security requirements, and reduces the range of tools needed. Our initial tests show good performance characteristics, and Data Source Providers have been updated to be compatible with Snowflake ingest storage.

Until this release, TDI had a reduced set of ETL features for Snowflake Prepare; with this release, almost all of the missing features become available. It's now also possible to use Add Related Records, Integrate Existing Objects, and pre- and post-scripts with stored procedures. In the latter option, you can also call engines other than Snowflake SQL, such as Python, without extra configuration. With some adjustments in Snowflake, you can also integrate Java, JavaScript, and Scala into your solution. The only feature still on the roadmap is Object Level Security and its derived options.

With this expansion of capabilities, switching from SQL Server to Snowflake (and vice versa) is now as easy as it can be. There are still some subtleties around the differences between Snowflake SQL and T-SQL and data types, but this provides a lot of extra flexibility.

MCP Server

If you've been following AI news, you'll have seen several MCP server releases pass by recently. MCP servers standardize the availability of capabilities that AI models can use. In the context of TimeXtender, you can now make the structure and meaning of Deliver instances (data marts) and the data within them accessible to the AI models and agents you want to use in your organization. With the MCP Server, AI can “understand” how your data models are put together, and results can actually be extracted from them. Because TimeXtender is metadata-driven, and that metadata feeds the MCP Server, this gives you a fantastic foundation for better outcomes from AI.

The server has an integrated connection with Claude Desktop and can also easily work with other models, including on-prem models from, for example, LM Studio. This gives you the freedom to put together an AI solution that fits your organization's policies. Of course, running AI on-prem does require solid hardware and expertise.

Questions covered by your model can be analyzed in depth.

Although this functionality seems very simple, it opens up a world of possibilities: self-service analytics is now within everyone's reach. Our own tests against AdventureWorks data show that the more advanced models can produce in-depth analyses using the data models you provide yourself. Of course, this does require well-defined models and good data quality. More on that later.

Of course, this also means “With great power comes great responsibility”: to work with AI in a sustainable way, you need a policy framework and the awareness that you can't always blindly trust AI. With a well-designed data model, equipped with the right descriptions, you can get a lot more value out of the data you already have.

When you phrase the question in exactly the right way, you can trust the result. We'll follow up on this topic with a more detailed blog later.

For more details, see here.

My colleague Louis de Roo will soon be giving a KPI Masterclass (February 13) where you can learn how to better define your metrics. Check out our Academy agenda here.

Prepare-to-Prepare data movement

From now on, it's possible in TDI to move data from one Prepare instance to another. This lets you build more complex implementations, such as a situation where you want to facilitate a group of separate data warehouses for business units, along with an overarching data warehouse that brings everything together. You can also use this to split your solution into smaller pieces to optimize your infrastructure. Currently, this works for SQL-based instances.

Salesforce Connector

Salesforce is a product that can be difficult to unlock with standard connectors like the REST connector. From now on, there's also a dedicated Salesforce connector that has extra filters to only pull in the most commonly used tables, or to take over all table and column names in the “friendly” variant.

For more details, see here.

Orchestration Execution Error Insights with XPilot

When you reload your data warehouse, you sometimes run into errors. This is often accompanied by a hefty chunk of logging and stack traces. Since it can be difficult to extract the right information from that, Orchestration now has XPilot-enhanced error insights: in short, AI is used to derive the likely issue and the steps to resolve it, in readable text, from the various logs. This makes it much faster to get straight to the core of the issues and prevents you from going down the wrong path.

For more details, see here.

Data Quality Rule Generation with XPilot

In Data Quality, you can now have XPilot analyze your datasets to automatically generate data quality rules. You can steer XPilot in more detail and tweak and adjust the generated rules.

For more details, see here.

Qlik Cloud

“But wait, there is more…”: Qlik Cloud now also has its own Deliver Endpoint, with which you can push your models as apps to Qlik Cloud and reload them. This eliminates the need for the workarounds that were previously necessary. This works both for a direct connection to your data and via the Qlik Data Gateway. This makes my (former) green heart beat a little faster again…

At the moment, Qlik Cloud Managed Spaces aren't directly supported yet; feel free to get in touch with our own Qlik Ambassador Lennaert van den Brink if you'd like to know more about Qlik.

Conclusion

This release brings a whole lot of great features that, especially in the area of AI, draw on the core of what makes TimeXtender a great product: a metadata-driven, end-to-end data solution.

Written by Ruairidh Smith,
Senior Consultant at E-mergo