Blog

Introducing Data MCP

By Suresh Thalluri

Headlines keep surfacing about stalled AI projects. Atlassian reports only 4% of organizations see meaningful ROI from AI, while a recent MIT study found 95% of generative AI pilots fail to reach production.

We’ve seen firsthand why this can happen inside enterprise teams through our agentic AI development work at Lean Innovation Labs.

The problem we kept seeing

Consistently, one of the main places where we see AI pilots stall is in the integration work.

Connecting agents to live data in existing systems requires building custom middleware (MCP layers) for each source. Every connector adds weeks of development time and another costly deployment cycle. Then even after launch, teams are left with little visibility into how agents use the integrated tools and data, which makes it harder to build trust and expand agent adoption within their organization.

To help shorten that time to impact, we built Data MCP. And today, we’re releasing it as an open source project.

What Data MCP offers

Data MCP exposes databases and APIs as Model Context Protocol (MCP) tools. Instead of writing one-off middleware, teams can configure parameterized queries and allow agents to use them in plain language. Every tool call is logged, providing the transparency needed for governance and monitoring.

Key capabilities include:

  • Connections to PostgreSQL, MySQL, SQLite, and Databricks, with more integrations planned
  • Secure, reusable queries built with Jinja templating and parameter validation
  • A web interface for managing data sources, building tools, and reviewing activity

How you can put Data MCP to work

Curious how to use Data MCP inside your organization? Here are some example applications and use cases for tools you can build to expose useful data and smooth operations across teams:

  • AI Agents for Insurance Companies: An agent needs to check claims over $10K pending for more than 30 days. With Data MCP, you can build a query to retrieve the claims base and expose it as a tool to the agent.
  • AI Agents in Enterprise Sales: In a pipeline review, you want an agent to pull current deal data directly from the CRM database and highlight accounts at risk. Use Data MCP to build a tool that retrieves the CRM data and deploy it rapidly.
  • AI Agents in Customer Support Ops: A support bot queries logs to surface live incidents, reducing manual triage using the tool provided with Data MCP.

These examples show how connecting agents to live data helps pilots move past proofs of concept and start delivering measurable impact to productivity.

What’s new in this release

  • A UI-driven tool builder for faster setup
  • More powerful templating for dynamic queries
  • Monitoring features that capture query execution details
  • Expanded data connectors across databases and APIs
  • Streamlined onboarding to get projects moving quickly

Why open source

This project reflects what we have learned from deploying agentic frameworks in real enterprise and government environments. By sharing it as open source, we want to shorten the middleware cycle for others and help teams reach impact sooner.

Get started

Data MCP is available now at github.com/leaninnovationlabs/dmcp. Setup takes about ten minutes, and you can connect your first data source immediately.

If you find it useful, please ⭐ the repo. It helps others discover the project and gives us feedback on where to take it next.