MCP in the Wild: CData's First 6 Weeks of Customer Adoption

When Anthropic introduced the Model Context Protocol (MCP), it promised to revolutionize how AI assistants interact with data. Six weeks after releasing our beta CData MCP Servers, the numbers tell a story that even we didn't expect.

Explosive growth in just 6 weeks

The trajectory has been nothing short of remarkable. What started as 12 downloads in the last week of April exploded to 81 downloads in the second week of June, speaking to the massive pent-up demand for AI data connectivity.

The acceleration has been particularly dramatic in recent weeks, with the past two weeks alone accounting for 187 downloads—nearly two-thirds of our total adoption in just the final stretch of our measurement period.

In total, we've seen 294 downloads across our MCP server portfolio, representing 255 unique users from 139 different companies. From Fortune 500 enterprises to innovative startups, organizations across every industry are exploring connecting their LLMs to their enterprise data sources.

The data sources driving adoption

The download patterns reveal fascinating insights about which data sources are initial targets for AI connectivity.

The clear leaders:

• Google Sheets (23 downloads): No surprise here. Sheets remains the universal data collaboration platform
• NetSuite (17 downloads): Enterprise ERP data – especially in difficult to access systems - is clearly of high interest
• QuickBooks Online (12 downloads): SMBs using AI to understand financial data
• SQL Server (12 downloads): Traditional databases exposed through modern chat interfaces
• QuickBooks (11 downloads): A close follow to its online counterpart
• Salesforce (11 downloads): CRM data accessible through LLM is insanely powerful (and a personal favorite source of mine)
• Zoho CRM (11 downloads): The SMB CRM market is equally hungry for AI integration

The rapid adopters who’ve gained the most traction in the past two weeks:

• Excel/Excel Online (18 combined downloads): Microsoft's spreadsheet empire looking to MCP connectivity
• Google Analytics (8 downloads): Another difficult to access and understand system made available through conversational interface
• Power BI XMLA (8 downloads): Business intelligence meets AI assistants
• MySQL (7 downloads): Open-source databases leading the charge
• SAP/SAP HANA (14 combined downloads): Enterprise systems available through AI

From downloads to real usage: The production story

Downloads are exciting, but actual usage tells the real story. Our telemetry data from the first 6 weeks shows customers aren't just experimenting—they're incorporating it into their regular routines:

Scale that surprised us

• 26,475 total queries across all MCP Servers
• 2 million rows of data processed
• 188 unique deployment nodes (individual server instances running across different environments, teams, and use cases)
• 57 different data sources actively used

The workload patterns emerge

The usage data reveals three distinct patterns emerging:

1. Enterprise CRM powerhouses Salesforce leads with 842,028 rows read across 52 unique deployments, generating 7,332 total queries. This isn't just testing— this is users who’ve realized how much easier and faster it is to get their questions answered than building out reports or dashboards.
2. Database integration champions SQL Server deployments show the most write activity (524 write queries), indicating customers aren’t only reading data but also updating databases from their AI assistant.
3. Cloud-native analytics surge Google Sheets (37 unique deployments), Google BigQuery (370,019 rows processed), and Azure DevOps (544,240 rows read) show the ease of connecting cloud sources to AI assistants with CData MCP Servers.

The enterprise signal is loud and clear

Looking beyond the raw numbers, the quality of adoption is what excites us most:

• 188 unique node IDs suggest significant enterprise deployment across multiple environments
• Read-heavy patterns (25,204 read vs 1,271 write queries) reveal a unique strength of CData’s MCP Servers
• Mixed traditional and modern data sources show organizations want to connect their entire data ecosystem to AI

The read-heavy revolution

The 20:1 ratio of read-to-write queries tells a compelling story about the CData MCP Server advantage. Unlike MCP Servers buit on limited API integrations, CData MCP servers are built on CData’s industry leading connectivity. Meaning our MCP Servers excel at deep data exploration and real-time analysis.

This read-heavy pattern also reflects AI's natural strength: rapidly consuming, analyzing, and synthesizing information from multiple sources to generate insights. LLMs paired with CData MCP Servers are uniquely positioned for these use cases, leveraging our 10+ years of data connectivity expertise to provide:

• Enterprise-grade data access with full SQL query capabilities, not just simple API calls
• Real-time querying with optimized performance across diverse data sources
• Advanced filtering and aggregation at the source level, reducing data transfer overhead
• Data in-place access that doesn’t require data to be moved or replicated; preserving schema, metadata, and user permissions

While automation-focused MCP approaches excel at specific tasks, CData MCP Servers deliver the depth and performance needed for complex analytical workloads. Our servers don't just connect to data—they unlock its full analytical potential for AI-driven insights.

The fact that we're seeing everything from traditional databases (MySQL, SQL Server) to modern cloud services (Snowflake, BigQuery) to business applications (Jira, Confluence) tells us something profound: Users want it to be easier to interact with their data and they're turning to AI to do it. 

Where we go next

These first 6 weeks provided a fascinating glimpse into how users want to connect and use their enterprise data within the AI context:

The CData MCP advantage is real

MCP Servers built on traditional API integrations are often limited by the underlying integration’s capabilities. CData MCP servers provide instant, conversational access to data sources with deep connectivity. The rapid adoption curve shows enterprise users immediately recognizing this value.

Enterprise-grade from day one

We expected more testing and early exploration with our beta MCP Servers - but we're seeing serious workloads. When Salesforce installations are processing 842K rows and SQL Server deployments are handling bidirectional data flows, that's meaningful usage.

The democratization effect

The diversity of data sources—from enterprise SAP HANA to simple CSV files—shows that MCP is democratizing AI-data access to any enterprise data source. Giving users access to these sources via an LLM interface allows users to explore data without intermediary steps like extensive data centralization or report or dashboard building. In many ways fulfilling the long awaited promise of data democratization efforts.

Looking ahead: The 10x opportunity

If the first 6 weeks taught us anything, it's that we've only scratched the surface. The growth trajectory suggests we're witnessing the early stages of a fundamental shift in how organizations think about data access in the AI era.

The next 6 weeks will be even more interesting. As word spreads and more organizations deploy MCP servers into production, we expect to see:

• Even more diverse data source adoption
• Larger-scale and centralized enterprise deployments
• The emergence of AI-native data architectures
• New use cases we haven't imagined yet

Get started today

Ready to join the MCP revolution? Our library of MCP Servers covers 350+ data sources. Whether you're looking to connect Salesforce to Claude, give your AI assistant access to your MySQL database, or enable natural language queries against your Google Analytics data, we've got you covered.

Download any of our MCP servers and join the 282 organizations already benefiting from conversational access to their enterprise data.

The revolution is just getting started—and these first 6 weeks are proof that the future of data access is conversational, AI-native, and happening right now.

Want to explore our full library of MCP Servers? Visit cdata.com/solutions/mcp to see all available connectors and get started in minutes.