AI assistants (MCP)

lakelogic-mcp is an MCP (Model Context Protocol) server that lets an AI assistant, such as Claude Code, Claude Desktop or Cursor, check the contracts it writes with LakeLogic itself, rather than guess. It runs on your machine, needs no LakeLogic account, and reads only the project folder it was started in.

Install

claude mcp add lakelogic -- uvx lakelogic-mcp

Any other MCP client: run uvx lakelogic-mcp as a stdio server. --root <folder> serves a folder other than the current one.

Tools

Tool	What it does
`validate_contract`	Is this a valid Open Lakehouse Contract? Each error's location and problem
`lint_contract`	Is it a good one? Governance findings from `lakelogic lint`, each with its fix
`explain_rule`	Why a lint rule exists: what it checks, why it matters, a failing and a fixed contract
`contract_reference`	How one part of a contract is written (`quality`, `materialization`, ...)
`export_odcs`	The contract as an Open Data Contract Standard document
`generate_test_data`	Synthetic data shaped by the contract, with a share of rows that deliberately break its rules
`run_contract`	Runs the contract on a local CSV or Parquet file with Polars or DuckDB: rows passed, rows quarantined, and why. Nothing is written to the contract's target.

Example

Generate 200 rows of test data for contracts/silver_trips.yaml with 15% invalid rows, run the contract on it with Polars and DuckDB, and tell me what was quarantined and why.

The assistant writes the data, runs the contract on both engines and reports the good and quarantined counts and each rejection reason.

On the LakeLogic platform

With a LakeLogic account, an assistant can also work with your governed estate: assess a contract's quality gaps, propose a rule, test it in a sandbox, and have a person approve it, with the evidence recorded. See Connect your AI assistant (MCP) on the platform docs.

Source: github.com/LakeLogic/lakelogic-mcp.