Overview
Great Expectations MCP Server bridges Great Expectations data-quality checks to the Model Context Protocol (MCP), enabling AI agents to interact with data validation tooling via MCP transports. It lets you programmatically load datasets from CSVs, URLs, or inline data; connect to Snowflake and BigQuery via URI prefixes; define and modify ExpectationSuites (profiler flag deprecated); run validations and retrieve detailed results, either synchronously or asynchronously. The server supports multiple transport modes (STDIO, HTTP, Inspector GUI) and can store datasets and results in-memory by default or SQLite for persistence. It offers optional authentication for HTTP clients (Basic or Bearer), per-minute rate limiting, and origin restrictions. Prometheus metrics and OpenTelemetry tracing are available to integrate with observability stacks. The server can be run via Docker or in local development, and is designed to be controlled by MCP clients (e.g., Claude) to load data, apply expectations, execute checks, and feed structured validation outcomes into larger agent workflows.
Features
Load CSV data from file, URL, or inline
Load CSV data from various sources, including files, URLs, or inline data, up to 1 GB (configurable).
Load tables from Snowflake or BigQuery using URI prefixes
Connect to Snowflake or BigQuery via URI prefixes and load tables as datasets.
Define and modify ExpectationSuites (profiler flag deprecated)
Create, update, and manage ExpectationSuites; note that the profiler flag is deprecated.
Validate data and fetch detailed results (sync or async)
Run validations and retrieve detailed results, available in synchronous or asynchronous modes.
Storage options: in-memory or SQLite
Store datasets and results in-memory by default or persist to SQLite for durability.
HTTP authentication for clients
Support Basic or Bearer authentication for HTTP clients to secure access.
HTTP rate limiting and allowed origins
Configure per-minute rate limits and restrict origins to control access.
Observability and transport modes
Prometheus metrics, OpenTelemetry tracing, and support for STDIO, HTTP, and Inspector GUI transports.
Who Is This For?
- LLM agents:Interact with Great Expectations checks via MCP to load datasets, apply expectations, run validations, and fetch structured results for downstream reasoning.
- Data engineers / AI developers:Expose GE data-quality tooling to MCP-enabled applications and automated workflows to validate data within pipelines.
