AI is transforming how teams work with data, but without the right context, even the best models fall short. That’s why we’re introducing support for the Model Context Protocol (MCP), an open standard that defines how AI systems interact with structured metadata in a secure and consistent way.

With MCP, your existing AI tools like Claude and Cursor can connect to your Secoda catalog and access trusted metadata like lineage, glossary terms, documentation, and SQL context, right where you work.

What is MCP?

MCP is an open protocol that standardizes how external AI tools and agents can securely interact with metadata systems like Secoda. Think of MCP like HTTP for AI. Just as HTTP offers a protocol for communication on the internet, MCP provides a consistent method for tools like Cursor, Claude, or custom agents to access information like lineage, documentation, glossary definitions, and other metadata from Secoda.

MCP uses a client-server architecture. AI tools connect to MCP server endpoints hosted by Secoda and are authenticated based on workspace-level permissions. Once connected, these tools can search your catalog, retrieve documentation, run SQL queries, and explore lineage. They also respect your access controls and governance policies.

Why MCP?

Organizations today have a rich layer of metadata spread across dashboards, warehouses, and documentation. But that context is often locked inside individual tools. MCP makes it possible to bring that context into the AI tools your teams already use.

By connecting to Secoda’s MCP server, your AI tools can:

• Search for tables, dashboards, columns, and documentation
• Understand lineage and data relationships to answer impact questions
• Access your business glossary for consistent terminology
• Execute SQL queries using metadata context
• Respect all workspace roles and permissions for secure access
• Use a standardized interface to ensure consistency across tools

MCP makes it possible to embed trusted metadata into any AI workflow, extending the reach of your data catalog without duplicating logic or breaking governance.

What MCP enables in Secoda

Once connected to Secoda’s MCP server, external AI tools can:

• Search your data catalog for assets, definitions, and documentation
• Retrieve metadata about tables, columns, dashboards, and metrics
• Explore upstream and downstream lineage between data entities
• Execute SQL queries directly on your connected data warehouse
• Access glossary definitions and knowledge articles
• Retrieve context-aware answers using governed metadata

This enables your AI tools to act with the same context and confidence as your data team, without ever leaving their native environments.

Once MCP is connected, you can use natural language to interact with your Secoda catalog. Test it out by asking questions like:

• “Find all tables related to customer data”
• “What would be affected if I change the orders table?”
• “Show me the schema for the sales_summary table”
• “What does ‘MRR’ mean in our glossary?”
• “Run SQL to get average order value by region”

Use cases for teams

Here’s how teams are already seeing value from using MCP to bring Secoda’s trusted metadata into their AI tools:

• Analysts can ask complex questions and get structured responses with charts, queries, and documentation in a single thread
• Data engineers can trace the impact of schema changes or pipeline failures by asking about upstream and downstream dependencies
• Governance leads can ensure AI answers are consistent with glossary definitions and don’t expose sensitive assets
• Platform teams can integrate their own tools with Secoda through MCP, enabling custom workflows or internal agents
• Executives and business users can ask natural language questions and receive answers that reference trusted, internal sources

MCP enables AI systems to act with the same context, standards, and structure your team relies on.

Security and governance built in

Every operation via MCP follows your existing Secoda governance framework. AI tools only access metadata that the authenticated user or service is authorized to see, and all interactions are logged for full traceability.

With MCP, you get:

• Fine-grained access control, enforced across AI-generated outputs
• Consistent behavior across different AI tools, regardless of vendor or model
• Seamless integration, without complex configuration or manual setup

This means your organization can scale its use of AI with confidence, knowing that governance isn’t sacrificed for speed.

How to enable MCP in Secoda

If you're already using Secoda AI, MCP is already live in your workspace. You don't need to take any additional steps to activate it.

When an AI tool connects to Secoda:

• It authenticates through your MCP server endpoint
• Access is granted based on your workspace’s existing permissions and governance policies
• All operations—whether retrieving documentation, running SQL, or exploring lineage—respect the security and access controls you’ve configured

To use MCP, all you need is an active Secoda workspace with AI features enabled and an MCP-compatible AI assistant (like Claude, Cursor, VS Code, etc.).

Want to connect to a specific tool like Claude or VS Code? Check out our detailed setup guide.

Looking ahead

The future of AI in data isn’t just about smarter models. It’s about structured context, open standards, and systems that reflect how teams actually work. MCP is a major step in that direction, not only for Secoda but for the broader ecosystem of AI applications.

To learn more about the open standard behind MCP, visit modelcontextprotocol.io, or reach out to our team to see how you can get started.