Two discovery tools so the model stops guessing your schema.

The scope is narrow and stated plainly: an assistant that can list your tables and read their columns writes far fewer invalid queries. Everything past discovery still runs through your own code.

What it is

A server built into the Intugle library. Run intugle-mcp from your project root and it serves your generated semantic model at http://localhost:8080/semantic_layer/mcp over HTTP.

What you get

get_tables returns every table in the semantic model with the technical description attached to it
get_schema takes a list of table names and returns their columns, data types and other metadata, including the links between tables
One URL any MCP client can take — the docs carry ready snippets for Cursor, VS Code, Gemini CLI, JetBrains AI Assistant and Claude Code

Requirements

A semantic model built with SemanticModel and loaded — the discovery tools only appear once one exists. The server runs on your own machine, so the data stays where it already is.

Setup effort

Follow the published steps — pip install intugle

See installation options

USE CASES

01

Getting an assistant oriented in an unfamiliar warehouse

Ask for the table list first — each table comes back with its description, so the assistant picks the right one before reading a single schema.

get_tablesget_schema

02

Writing a query against tables you have not memorised

Pass the table names to get_schema and the column names, types and inter-table links come back together.

get_schemaget_tables

03

Coding against your own data in the editor

Point Cursor, VS Code or Claude Code at the local endpoint and the assistant works from your semantic model instead of a guess.

get_tablesget_schema

2 tools

Listed in project docs. Your client may show a different list.

get_tables

get_schema

BEFORE YOU INSTALL

What to know first

Tools you get

2 tools named in the project's documentation.

Requires

pip on PATH
LLM_PROVIDER and your LLM provider's API key in your environment (the variable name follows the provider, e.g. OPENAI_API_KEY)
A semantic model built with SemanticModel in your project

Where it runs

All 5 documented clients run the server on your machine.

INSTALLATION

Connect your client

Server setup

TERMINAL

pip install intugle

Configure the LLM the library uses to build the model — LLM_PROVIDER plus your provider's key (the variable name follows the provider):

TERMINAL

export LLM_PROVIDER="openai:gpt-3.5-turbo"
export OPENAI_API_KEY="<YOUR_OPENAI_API_KEY>"

In Python, build the semantic model over your datasets with from intugle import SemanticModel, sm = SemanticModel(datasets, domain="Healthcare") and sm.build() — the project's quickstart notebooks walk through it.

Start the server from your project root (it listens on localhost:8080):

TERMINAL

intugle-mcp