CData Software Inc. has launched Connect AI Gateway, a platform designed to manage how AI agents access enterprise data, select models, and utilize tools. The product is currently available in early access and expands upon the company’s existing Connect AI platform and Model Context Protocol offerings.
The gateway provides a centralized interface for IT teams to register models and agents while establishing usage policies. It features a context engine that integrates business definitions from tools like Fivetran’s dbt and Microsoft’s Power BI, aiming to improve response accuracy and reduce data exposure.
The platform includes routing and control functions that allow companies to set token budgets and enforce access permissions at the row and column level. It also maintains an audit trail for every prompt and action, helping administrators monitor agent behavior and investigate potential errors.
CData reports that its data layer can filter and aggregate records before sending them to a model, which may lower token costs. In internal tests involving 378 enterprise queries, the company claimed a 98.5% accuracy rate, compared to 65% to 75% for other tested providers.
The gateway captures knowledge from documents and user interactions into a context graph, which remains independent of specific AI models. While the current release does not support direct imports from all data catalogs, the company plans to add this functionality in future updates.
