We're excited to share a significant enhancement to CData Connect AI, the managed MCP platform. As we continue to lead in innovation for MCP connectivity, we are releasing enhanced MCP instructions across 350+ enterprise systems. AI apps integrated to enterprise sources through Connect AI will work more intelligently and accurately with instructions that encode CData’s deep knowledge of every enterprise source system.
Source-specific MCP instructions: A guide to your data for AI
With today's enhancement, many connectors across our portfolio now include enhanced MCP instructions — a pre-configured guide to key systems that helps AI instantly understand data source schema, metadata, and entity relationships for more efficient data queries and token consumption.
Solving core LLM challenges with enterprise data
When AI interacts with enterprise data sources directly, it faces inherent limitations. Large language models struggle when presented with more than 50 objects without a clear hierarchy — it starts ignoring options and misses relevant data. AI is not naturally skilled at discovering complex relationships between entities, and when working with sources that have both static and dynamic schemas, it tends to prioritize pre-studied static objects over discovering what's actually available.
Left to its own devices, AI tends to favor quick responses over completeness. It often executes queries without fully understanding the schema, attempting to query tables or columns that don't exist. Filtering requires exact value matches that AI may not know. And when it comes to write operations, the margin for error shrinks dramatically — schema accuracy becomes critical.
These limitations lead to frustrating outcomes: AI confidently declaring "there are no such objects" or "that operation is not supported" when they do exist. AI may construct expensive queries that scan entire datasets, resulting in timeouts. Without proper guidance, AI wastes API calls and tokens making queries that fail simply because it didn't understand the schema.
How enhanced MCP instructions increase AI understanding and efficiency
CData's source-specific MCP instructions address these challenges by providing AI with the guidance it needs before it ever starts exploring data.
Rather than overwhelming AI with too many unknown objects to choose from, MCP instructions surface and map core entities, giving AI a strategic starting point instead of a sprawling catalog. They provide step-by-step guidance for dynamic schema discovery, helping AI understand how to find additional objects when needed rather than assuming they don't exist.
The instructions include progressive query guidance, walking AI through approaches from simple lookups to complex joins across related entities. Sample stored procedures demonstrate common operations, while data model and relationship guides help AI understand how entities connect — turning it from a data novice into a knowledgeable analyst from the first interaction.
Built on decades of connector expertise
CData's MCP instructions encode deep, source-specific knowledge from 10+ years of building enterprise connectors. For example, when AI connects to Salesforce through Connect AI, it understands Account-to-Opportunity relationships as well as any custom fields in your environment. For NetSuite, it knows multi-subsidiary structures. For ServiceNow, it recognizes how incidents relate to change requests.
AI leverages these source-specific instructions to query like a domain expert from the first interaction, whether you're building conversational analytics for business users or deploying autonomous agents in workflows connected to enterprise data.
Data context for AI that scales
This enhancement further strengthens CData's context-aware approach to MCP, giving AI he context from your business it needs to reason like expert human analysts:
Source-level semantic intelligence – AI understands the business meaning behind data through metadata context, enabling more accurate reasoning with human-like understanding.
Lean token footprint – Query push-down and derived views keep AI focused on reasoning rather than data exploration, delivering more predictable costs and faster responses.
Curated data collections and tools – Create targeted, multi-source datasets with custom tools to direct agents for specific use cases—from financial analysis to customer service.
Documents as data – Retrieve and edit files directly through your AI workflows without complex RAG pipelines.
Start using real-time business context with AI
New source-specific MCP instructions in Connect AI's managed MCP platform enable AI to better understand and interact with your business context. Instead of learning your data structure through trial and error — burning costly tokens and consumption along the way — AI assistants and agents start with built-in intelligence about connected systems at the first query, resulting in more accurate insights, lower overall token costs, and faster time to value. CData has honed this connector-specific expertise now encoded for AI in MCP instructions over the past 10+ years, used by leading enterprises like Palantir, Salesforce, and Google.
Whether you're building conversational AI for natural language analytics or deploying autonomous AI agents, CData Connect AI delivers the context your AI needs to work smarter from day one.
Start trying it yourself in minutes with a 14-day free trial of Connect AI.
Explore CData Connect AI today
See how Connect AI excels at streamlining business processes for real-time insights.
