The rise of Agentic AI and Large Language Models (LLMs) is transforming classic data integration, with the Model Context Protocol (MCP) emerging as a key piece of the new tooling. This protocol is changing how users interact with model data, and traditional data companies now race to meet the new requirements.
ISVs and enterprises can’t move fast enough on the new AI front, and traditional business databases will often be central. BI reporting will be an early target. Software architecture leads may turn increasingly to data connectivity providers like CData Software if they are going to move fast without breaking things.
For its part, CData has moved quickly to address the rising market. This week the company announced its Connect AI software, which taps MCP to integrate AI assistants, workflow builders, agent platforms and frameworks with over 350 enterprise data sources, giving individuals a single place to manage connections to different LLMs and frameworks. Connect AI follows CData’s roll out this Spring of CData MCP Servers.
Track Record in Connectivity
Established as CData Software in 2014, the company has a long track record of expertise in connectors and drivers that bridge diverse formats. The tooling works with APIs to extract, transfer and transform different pools of data.
ODBC and JDBC are prime examples of bridging components that create standardized connections to MySQL, SaaS applications like Workday and Salesforce and many more.
The company counts Palantir, Google Cloud. SAP and Salesforce among companies that have embedded CData technology into their offerings.
CData added advanced data virtualization capabilities last year, with the acquisition of Germany-based Data Virtuality. Such capabilities, which dramatically reduce requirements for data movement, could well come to play as company addresses AI processing.
MCP connection
Ahead of the Connect AI announcement, we spoke with CData’s Manish Patel, Chief Product Officer and Will Davis, CMO. Patel told me that CData views MCP as an addition to its portfolio — as a new way for customers to apply their existing connectivity capabilities, but with a new AI-driven focus.
Will Davis said CData’s implementations prioritize efficient query processing while improving governance for connectivity. To that end, connectors improve AI model security, with data access logged under the identity of authenticated users or agents.
CData’s Davis explained the nature of the fit: ‘We have built out a connector library of 350-plus sources, and realized when the Model Context Protocol came out, that we could utilize [MCP] to make the data sources for which we provide connections more available to LLMs and different AI or agentic frameworks.”
Mind the tokens
Now in open-source distribution, the MCP API was first created by noted AI-start-up Anthropic. Its goal was to connect different LLMs to different tools or data sources, and a good portion of Agentic AI efforts are now riding on the rise of MCP use.
There are plenty of steps involved along the way, steps that can benefit from long-time data integrators’ know-how, Manish Patel suggested. An example is query handling that optimize for minimal LLM token creation.
“Our software is intelligent enough to know how much of a query can best be pushed down server. If there’s aggregation or any sort of predicates that need to happen, the majority of the time we’re pushing that down to the server side,” he said. “So, we’re not pulling large data sets back into the LLM for it to then have to do aggregation or grouping or field covering.”
That’s a big advantage, he said, in controlling token consumption, which viewers agree will become more and more important.
The New Importance of “Plumbing”
Of course, moving MCP forward will require its share of such efforts, but it appears to be a suitable foundation.
According to Patel, while there are refinements and capabilities that need to be added as it gets adopted, “from a core-principle perspective, Anthropic was not far off the mark in creating connectivity to LLMs.”
“It pretty much does that out of the box,” he said.
Often described as ‘plumbing,’ the data connectors from companies like CData are taking on a new level of importance.
While the term might have a workmanlike connotation, this vital ‘plumbing’ is far from everyday common. It is central to modern application design. Solutions like CData Connect AI usefully expand the capabilities for implementors.
Added intelligence in MCP-based connectors could provide better answers to AI’s many questions. Ω
This also appeared on Medium.com and LinkedIn.