Build agents, copilots, and reports on live enterprise data

One governed data layer across hundreds of enterprise systems. Your agents, copilots, and dashboards reach live data in real time, with no data movement and no connectors to build.

0%
Accuracy across 378 real-world prompts
0%
Token reduction on multi-source queries
Hundreds
Of live enterprise sources
30 min
To first query
Managed

CData Connect AI

CData runs the connectivity for you, hosted and governed. MCP, ODBC, JDBC, SQL Server, REST — one endpoint for everything your team is building.

Read the Docs
Self-hosted

CData Drivers

Same connectivity, running in your own environment. ODBC, JDBC, ADO.NET, Python, and SQL Server interfaces. Data stays on your network.

Jump to Drivers

One interface. Every protocol your consumers already speak.

Connect AI exposes your enterprise data over MCP for AI agents, ODBC/JDBC and a virtual SQL Server endpoint for BI tools, and REST and OData for custom apps. You connect a source once, set access, and every consumer queries it live. CData maintains the connectors so you don't have to.

Managed

CData hosted and maintained, no connectors to build or API changes to manage.

Governed

Source permissions enforced at query time, full audit logs, SSO.

Security overview

Live

Queries run in place against the real source. No copies, no sync jobs.

Governance is the hard part.

Every vendor now ships an MCP server. The trouble starts when every team runs their own, with separate auth and no central control. That's how agent projects stall before production.

Connect AI puts one governed data layer in front of all of it. Access enforced at query time. Every interaction logged. See the benchmark results.

What are you building?

Pick the surface closest to what you're shipping today.

AI agent 10 mins

Scaffold an agent on live data

Stand up a working agent wired to Connect AI's MCP endpoint, then customize. Point any MCP client at Connect AI and it can list tools and query live data.

curl -X POST https://mcp.cloud.cdata.com/ \
  -H "Authorization: Basic $CDATA_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"jsonrpc":"2.0","id":1,
"method":"tools/list"}'
By AI Framework
MCP Reference
AI assistant

Build with your coding assistant

The remote MCP server gives Claude Code, Cursor, and GitHub Copilot live access to your enterprise data, so the code your assistant writes runs against real schemas the first time.

mcp add
    --transport http
    --name connect-ai-mcp
    --url https://mcp.cloud.cdata.com/mcp
Claude Code
Cursor
GitHub Copilot
Reports and dashboards 5 mins

Build reports and dashboards on live data

Point Power BI, Tableau, or your notebook at Connect AI and report on live data across hundreds of sources. No exports, no refresh jobs to babysit.

Server: tds.cdata.com
Protocol: ODBC / SQL Server
Sources: Salesforce, SAP, + hundreds more
Auth: Bearer token
BI and reporting tools
Power BI
Tableau

Why Connect AI, not another MCP server

Anyone can point an agent at a single MCP server. The enterprise problem is doing it across hundreds of sources, for every team, without losing control of what AI can see and do. Connect AI adds three things on top of raw connectivity.

Connectivity

Connectivity Illustration

One layer across hundreds of governed sources, including data behind your firewall through Connect Gateway. No fleet of scattered servers to run.

Context

Context Illustration

Connectors carry CData's per-source expertise, so agents get schema and business meaning instead of raw fields. In a benchmark across 378 real-world prompts, Connect AI reached 98.5% accuracy, above every other MCP provider tested. Curated custom tools cut Claude token use by 97.6% on a multi-source query.

Control

Control Illustration

Every query inherits the source system's own permissions at runtime. Scoped tools limit what each agent can see and do. Every interaction is logged, and write-backs are reversible and traceable.

Trusted by 10,000+ customers
Published results from GSK, UiPath, Office Depot, & Nintex
SOC 2 Type II & ISO/IEC 27001 certified

Need to keep it in your own environment?

If your data can't leave the network, or you're embedding connectivity inside a product you ship, CData drivers run where you run. Standard ODBC, JDBC, ADO.NET, Python, and SQL Server interfaces, self-contained, with no managed service in the loop. It's the same connectivity underneath Connect AI, just self-hosted.

Keep exploring

References, community, and fundamentals. Where to go after the quickstart.

Reference Docs

Browse by language

Per-language references and developer guides for shipping with CData connectors.

Community

Get answers from CData devs

Ask questions, share implementation patterns, and get answers directly from CData engineers and staff.

Fundamentals

New to CData?

Learn how the unified SQL interface, relational models, and caching work.

Frequently asked questions

  • How fast is it? What about latency?
  • What happens when a source changes its API?
  • What data can CData see, and where does it go?
  • Does my source work?
  • Am I locked in?
  • Is there a free way to start?

Connect your first source in 5 minutes

Governed live data to your agents, dashboards, and apps, through one data layer you don't have to run.