MindsDB MCP. Run ML Predictions and Audit Models via SQL.

Listen up. This MindsDB MCP Server connects your AI client straight into a full ML database engine using standard SQL. You're not just running queries; you’re executing predictions, auditing models, and checking data sources—all from one conversation window. It lets you run complex commands like SELECT ... PREDICT or CREATE MODEL without ever touching the command line itself.

Running ML-Enabled Queries

You use execute_sql_query to run any SQL statement against MindsDB's elements. This tool handles more than just standard data retrieval; you can execute entire machine learning workflows, whether that means setting up a new predictive model using CREATE MODEL or immediately fetching predicted data points with SELECT ... PREDICT. If you need the system to perform a computation that involves both historical data and a future forecast, this is your endpoint.

It allows you to run the full lifecycle of an ML model right through your agent's command.

Understanding Your Data Landscape

Knowing what data feeds into your predictions is crucial. You use list_databases to get a list of every external database connected to your MindsDB instance, showing you exactly where your data boundaries are—stuff like Snowflake or PostgreSQL. To understand the complex transformation logic applied to raw data, check out list_views.

This tool lists virtual views, which aren't raw tables; they’re mapped structural layers that let you query complicated transformations as if they were simple proxy tables.

Managing and Auditing Models

Before you predict anything, you gotta know what predictors you've trained. list_models gives you an inventory of all the deployed AI prediction engines (the models) available in your current project. It lets you see which algorithms are ready for querying right out of the gate. If you need deep details on one specific predictor—like its training parameters or status—you call get_model.

This retrieves detailed information and access points for an already trained AI model. You can also check the overall system health by using get_status, which pings the cluster to return current version numbers and diagnostic specifications, confirming that the entire MindsDB environment is running properly.

By combining these tools, you don't just query data; you manage the entire predictive pipeline. You list available models with list_models then execute a prediction using execute_sql_query, all while keeping track of your connected sources via list_databases and verifying system health with get_status. It’s how you run enterprise-grade ML logic without ever leaving your chat client.