As business intelligence enters the agentic era, we are seeing a fundamental shift in how teams interact with data. Today, we are announcing Looker’s fully managed Model Context Protocol (MCP) server in Preview. This introduction bridges the gap between enterprise-grade semantic data and the rapidly expanding ecosystem of AI agents, making Looker the plug-and-play semantic brain for any data-driven application.
The Evolution of the Looker MCP Ecosystem
Over the last year we have seen strong demand for the Looker MCP server through adoption of the open-source MCP Toolbox for Databases. Through collaboration with members of our BI Customer Council and early adopters, we saw how powerful it is to give AI agents access to Looker’s trusted semantic model.
Customers built incredible proofs of concept, most notably using AI to automatically generate and tweak complex dashboards on the fly from simple natural language prompts. However, running a local or self-hosted middleware server can create friction for enterprise adoption, as it requires dedicated developer resources to provision infrastructure, manage API keys, and maintain servers. Our new managed MCP server removes this friction. By embedding the MCP server directly into the Looker platform, we have turned multi-hour infrastructure configuration into a secure, zero-config connection that can take seconds.
Why Integrated MCP Matters for the Enterprise
By hosting the MCP server directly within Looker’s infrastructure, we unlock a secure and highly governed AI-ready data stack.
-
No deployment overhead: Hosted customers do not need to deploy containers or manage middleware. The MCP endpoint is natively exposed directly via your existing Looker instance URL.
-
Inherited security and VPC-SC compliance: The managed endpoint respects your existing VPC Service Controls and regional data footprints as long as your Looker instance is deployed within your boundary.
-
Seamless, multi-user identity: The managed MCP server maps agent actions to actual Looker users through standard OAuth2 flows. It works with your existing Looker SSO setup, whether you use Google Auth or a third-party identity provider.
-
Enterprise governance and auditing: Administrators gain full visibility. All queries executed via MCP will appear in Looker’s System Activity explores as API queries associated with the user whose credentials are used by the MCP server.
Driving Real-World Value: Enterprise AI-BI Use Cases
By exposing Looker’s semantic layer to LLMs, organizations can build custom agents while bypassing classic pitfalls of AI-driven data analysis, including hallucinated metrics or insecure database access.
Here is how the managed MCP server can deliver tangible business value:
-
Customized chat agents: Business users can securely chat with data directly from within their preferred AI assistant.
-
Dashboard Generation & Customization: Agents can programmatically generate and edit Looker dashboard tiles based on natural language commands.
-
Agentic Development & Semantic Modeling: Developers can integrate AI-assisted coding environments directly with Looker’s modeling layer to accelerate LookML development.
Get Started Today
Looker’s new fully managed MCP server is available for all Looker-hosted instances. Review our public documentation to learn more about how you can extend the power of Looker’s semantic layer to agents throughout your organization.