Most enterprise AI projects don’t fail because of the model. They fail because getting clean, governed access to production data is slow, political, and manual. Every new data connection requires IT involvement. Every new agent adds more surface area to manage. And every developer who goes around IT creates a new problem for someone else to clean up later.
CData is trying to fix that with three new releases aimed directly at developers: Connect AI Developer Edition (free), the CData Connect AI Python SDK (open source), and CData CLI.
The announcement, set for June 23, reflects a deliberate shift in how CData positions its platform. The company has been building enterprise data connectivity since 2014, starting with a driver catalog that let developers query Salesforce, Workday, Oracle, and hundreds of other systems through standard interfaces like JDBC and ODBC. Connect AI is the modern version of that same core capability, packaged as a SaaS product with MCP support, governance controls, and now a free entry point for individual developers.
The Governance Problem Developers Keep Running Into
When developers build AI apps on enterprise data today, they typically hit the same set of problems: authentication that breaks when tokens refresh, APIs that change without warning, rate limits that agents blow past without knowing, schema changes that make agent output silently incomplete, and pagination that the model has to manage itself — or doesn’t.
Each of these is manageable when you’re building a small internal tool. But in an agentic context, the problems multiply. You’re not managing one user’s authentication refresh. You’re managing dozens of agents, each with their own service identities, each hitting APIs that may have changed since yesterday.
“Our connectivity layer manages authentication per user and service accounts per agent,” said Jared [CData], who walked through the platform during a pre-launch briefing. “We automatically refresh tokens, and engineers don’t have to build that. They can just build whatever features they need.”
The same logic applies to schema drift. When a business user adds a custom field in Salesforce, an agent running a report on Monday morning won’t know about it on Tuesday. CData’s platform performs a metadata ping every time an agent hits the system, so schema discovery is dynamic. The agent sees what’s actually in the system right now, not what was there when the connection was built.
Rate limits are another common point of failure. When an agent is unaware of API rate limits, it will just hammer the endpoint — blowing up token spend or locking down the instance entirely. CData builds rate limit awareness directly into its connectors and handles backoff intelligently.
What IT Wants, and Why Regulated Industries are Coming First
The governance angle isn’t just a developer story. IT teams are increasingly pushing for a platform like this because they know developers will find ways to access data regardless. The question is whether they can provide infrastructure that makes it safe to do so.
“A lot of IT teams come to us with this idea in mind,” said Marie [CData]. “They’re like, ‘I know people are going wild out there, and I want to make sure I’m providing the infrastructure they can go wild on. So I feel like I’m covered, the organization is covered, I have the controls I need, but I’m not the barrier.’”
Counterintuitively, regulated industries — healthcare in particular — are among the earliest adopters. They can’t stop the train, but they want to make sure it’s running on a governed track.
Connect AI handles identity by authenticating a user or agent all the way back to that same identity in the source system. Toolkits let teams package governed data access into a single MCP Server URL, scoped to a specific use case, so agents have access only to what they need.
What’s in the Three Releases
Connect AI Developer Edition is free and includes the full enterprise feature set: MCP server support, per-user authentication passthrough, query logging with user-level attribution, and a management MCP server. The management MCP server is particularly useful — developers can provision connections, test them, and list available data sources directly from the terminal or IDE without going through a UI. It works out of the box with Claude Code, Codex, Cursor, LangChain, and any other MCP-capable tool.
The free tier also supports Toolkits, which scope data access to exactly what a given agent or use case requires.
The CData Connect AI Python SDK provides DB-API-compliant access to the same platform, so developers can pull enterprise data into existing Python workflows using pandas, SQLAlchemy, or cursor-based queries without changing their code. It’s open source and MIT-licensed, available on GitHub now.
CData CLI is a command-line interface for CData’s JDBC, ODBC, Python, and ADO.NET connectors. It’s designed so coding assistants like Claude Code and Cursor can scaffold data connectivity directly from the terminal without digging through documentation. The initial release supports JDBC, ADO.NET, and Python; ODBC support is coming in future releases.
One of the things beta users flagged as most useful: you can build on the free Developer Edition, and when you’re ready to deploy, flip it to an enterprise subscription. Everything you built was already governed, already auditable, and ready to scale. No re-architecture required.
The Build-vs-Buy Question
With MCP widely supported and APIs broadly available, it’s tempting to just build your own connectors. And a lot of teams do. But according to a December 2025 CData study of more than 200 IT leaders, 53% of organizations are still building API integrations and pipelines in-house to get data to AI models. That’s development time that isn’t going toward new features.
“The need to accelerate development exists everywhere,” said Jared. “Product owners are saying, ‘You have an AI coding assistant, why isn’t this shipping faster?’ And people are still running into the problem — I can ship that feature, but I also have to build the connector that gets the data to it.”
All three products are available at cdata.com/developers starting June 23.