Model Context Protocol (MCP) Server
Model Context Protocol (MCP) Server

Governance, everywhere your AI works

Collibra's MCP Server brings enterprise data governance to any AI agent or assistant working in your environment. Connect Claude, ChatGPT, Databricks, Snowflake Cortex or any MCP-compatible tool to Collibra and let AI act on data and context you can trust.

Read the blog Request a demo

Governance where
your AI already works

Extend your governance platform into every AI workflow and agentic pipeline you run. Your AI agents and assistants pull from a single source of governed data and context, wherever they operate. Purpose-built for the AI era, it's governance that moves with your organization.

Meet users and agents where they are

Collibra MCP Server brings governed data, glossary terms and lineage directly into Claude, ChatGPT, Databricks and more: no context switching.

  • Supported everywhere
    Works with Claude, ChatGPT, Databricks, Snowflake Cortex and any MCP-compatible tool
  • Natural language first
    No APIs to learn. Ask in plain language and get answers from governed Collibra data
  • One governed source of truth
    Every answer draws from the same governed data, not a disconnected copy

Unlock lineage and context on demand

Trace upstream sources and downstream dependencies in one query. Get impact analysis before a change ships: no diagrams, no SQL.

  • Upstream and downstream lineage
    See where data originates and what depends on it, in plain language
  • Impact analysis before changes ship
    Ask what breaks before you make a change and get affected reports instantly
  • Technical and business lineage
    Technical lineage for engineers. Business lineage for analysts. Same interface

Build and enrich your data catalog with AI

AI can propose assets, glossary terms and quality rules within your governance framework. Human review stays in the loop.

  • Create and enrich assets
    AI drafts catalog assets, from business terms to data contracts, ready for review
  • AI-generated data quality rules
    Point an agent at a dataset and get tailored quality rules by dimension
  • Governed write access
    Every write uses Collibra's existing permissions. AI acts within governance, not around it

Shape it to your use case and process

Production-ready and built to extend. Deploy anywhere, scope tools by team, and build custom skills for your workflows

  • 25+ governance tools, out of the box
    Discover assets, explore lineage, manage classifications and enrich your catalog
  • Extensible for any workflow
    Add custom tools and integrate with your platforms without rebuilding your governance stack
  • Token-efficient by design
    Built to minimize token use so workflows stay fast and cost-effective at scale

  • Data Engineers: ship faster, break less

    Run impact analysis before a change ships. Trace root causes in seconds when something breaks. Ask 'what depends on this column?' and get a structured dependency report without navigating the catalog manually.

  • AI Agents: Act on governed data autonomously

    Give your AI agents access to trusted, governed lineage, classifications and catalog data as callable tools. Agents reason over real metadata, not stale documentation, and act within your governance framework by design.

  • Data Stewards: Scale governance across the org

    Delegate catalog enrichment to AI. Propose business glossary terms from websites, documents, or transcripts. Review AI-generated proposals before anything is created. Governance scales without headcount scaling with it.

  • Data Analysts: Trust the numbers you present

    Ask where any number in a report came from. Trace a metric back to its source of truth. Confirm transformations in plain language, without reading SQL or learning table names. Every answer is drawn from governed Collibra data.

Learn more about creating a comprehensive view of your data landscape

Documentation

Collibra MCP Server documentation

Complete setup guides, tool references, and configuration documentation for the Collibra MCP Server. Connect your first AI assistant in minutes.

Read the docs
Developer Resource

Collibra on the Databricks Marketplace

Collibra's MCP Server is the most popular AI governance MCP on Databricks Marketplace. Connect directly from your Databricks environment and get started today.

Visit the Marketplace
GitHub

Collibra MCP on GitHub (CHIP)

View on GitHub

Learn. Grow. Be inspired.

From expert insights to guided learning paths and in-depth product resources, we make it easy for every Data Citizen to use data.

Consistently recognized as a leader, setting the standard for what’s next
View all industry acknowledgements

Frequently asked questions

What is the Collibra Model Context Protocol (MCP) Server?

The Collibra MCP Server is a production-ready Model Context Protocol server that connects any MCP-compatible AI assistant, agent, or platform directly to your Collibra instance. It makes your governed catalog, business glossary, lineage, data classifications, and data contracts callable tools, so any person or AI working in Claude, ChatGPT, Databricks, Snowflake Cortex, or a custom agent can access trusted, governed data and context in plain language, without switching tools or rebuilding integrations. MCP is an open standard, originally published by Anthropic, that defines how AI models connect to external tools and data sources. Collibra's implementation extends that standard to make enterprise governance natively accessible to AI.

Which AI platforms and tools does Collibra MCP Server support?

Collibra's MCP Server works with any MCP-compatible client. Current and planned integrations include: Claude (claude.ai, Claude Desktop, Claude Code), ChatGPT and the OpenAI API, Databricks (listed in the Databricks Marketplace), Snowflake Cortex, Microsoft Copilot, GitHub Copilot, Cursor, VS Code with MCP support, and custom agent frameworks built on the MCP open standard. Because MCP is an open protocol, any LLM or agentic framework that implements the MCP client specification can connect to Collibra, including internally built AI systems. The ecosystem is growing rapidly, and Collibra actively maintains marketplace listings and certifications across major AI platforms.

Learn more about the differences between data catalogs and data marketplaces here.
What can the Collibra MCP Server read and write?

Collibra's MCP Server supports both read and write operations, all within your existing Collibra governance and permissions model. Read capabilities include discovering and searching data assets and business glossary terms, retrieving upstream and downstream lineage, exploring technical and business lineage graphs, getting semantic context for tables and columns, reading data contracts, and searching data classifications. Write capabilities include creating and editing assets of any type, enriching catalog entries, pushing and pulling data contract manifests, adding or removing data classification matches, and proposing new business glossary terms. Every action uses Collibra's existing permission model, so AI acts within your governance framework, not around it.

How is Collibra's MCP Server different from using the Collibra API?

The Collibra API is powerful but requires developers to know what endpoints to call, what parameters to pass, and how to interpret the response. It's designed for machines executing deterministic instructions.The Collibra MCP Server is designed for AI, both AI models and human users working through AI. Instead of calling specific endpoints, you describe what you need in natural language ('trace the upstream lineage of the revenue_fact table' or 'find certified data products using customer data'), and the MCP Server resolves that request against Collibra using the right tools in the right sequence. It also abstracts away technical complexity: fuzzy name matching, concept mapping, and response formatting are all handled by the server, so the AI, and the human behind it, gets a direct, usable answer rather than raw JSON to parse.

How do I get started with the Collibra MCP Server?

The Collibra MCP Server requires an active Collibra instance (cloud or on-premises) and an OAuth 2.0 application configured in your environment.
To connect:

  1. Obtain your Collibra instance URL (format: https://[your-instance].collibra.com)
  2. Set up an OAuth 2.0 application in Collibra (admin access required, or request one from your admin)
  3. Connect your MCP-compatible client (e.g., Claude Desktop, Databricks, VS Code) using your instance URL + /rest/mcp as the server endpoint
  4. Authenticate via OAuth and start querying your governed data in natural language

Full setup documentation is available at docs.collibra.com

The road to Data Confidence™ starts here.