Trilogy Studio Core

A open-source IDE for exploring Trilogy, an experiment in streamlined SQL.

Try a hosted studio instance here, or run locally in docker.

Supports

• DuckDB
• BigQuery
• Snowflake

Read more about Trilogy, the language here.

This Repo

This repo contains the studio frontend, a minimal FastAPI language server that powers language features, and an MCP server example that can run queries.

It also hosts a sibling app: explorer/, an AI-native desktop chat experience that reuses studio's lib/ primitives (chat, charts, tools, providers). See explorer/README.md for the plan and the principles that keep the two apps from drifting apart.

Docker

The suggested local execution option.

A docker container is available in the base repo. Image runs resolution service [FastAPI] + statically serves frontend.

No telemetry enabled by default.

Quick Start

From repo root:

docker build -t trilogy-studio:latest && docker run -p 8080:80 trilogy-studio:latest   

Powershell

docker build -t trilogy-studio:latest . ; docker run  -p 8080:80 trilogy-studio:latest   

Access on http://localhost:8080 (or alternative port used).

Creating Direct Links

You can create shareable links that automatically import dashboards or editors from model stores into Trilogy Studio and open the relevant file. This is useful for sharing pre-configured dashboards or example queries with others.

URL Format

Import links use the following URL structure:

https://trilogydata.dev/trilogy-studio-core/#import=<model-url>&assetType=<type>&assetName=<name>&modelName=<model>&connection=<connection-type>[&store=<store-url>]

Store

The store parameter is optional, if provided it will register the relevant store as a store in the browser UX (useful if the model will be refreshed or updated over time).

Parameters

Parameter	Required	Description	Example
import	Yes	URL to the model JSON file	https://example.com/models/my-model.json
assetType	Yes	Type of asset to import: dashboard, editor, or trilogy	dashboard
assetName	Yes	Name of the specific asset within the model	Sales Overview
modelName	Yes	Name for the model configuration	MyModel
connection	Yes	Connection type: duckdb, bigquery, or snowflake	duckdb
store	No	Base URL of the model store for auto-registration	https://example.com/store

Examples

Dashboard Link (DuckDB)

https://trilogydata.dev/trilogy-studio-core/#import=https://trilogy-public-models.s3.us-west-1.amazonaws.com/tpcds.json&assetType=dashboard&assetName=Sales%20Dashboard&modelName=TPCDSModel&connection=duckdb

Editor Link (DuckDB)

https://trilogydata.dev/trilogy-studio-core/#import=https://trilogy-public-models.s3.us-west-1.amazonaws.com/tpcds.json&assetType=editor&assetName=Customer%20Query&modelName=TPCDSModel&connection=duckdb

Dashboard Link with Store Registration

https://trilogydata.dev/trilogy-studio-core/#import=https://example.com/models/sales.json&assetType=dashboard&assetName=Q4%20Sales&modelName=SalesModel&connection=duckdb&store=https://example.com/models

Connection Types

Different connection types may require additional inputs to create the connection for the end user.

• BigQuery: Prompts for Google Cloud project ID to use for billing.
• Snowflake: Prompts for account, username, and private key

Flexible Visualization

Explore data with easy, interactive visuals. Connect them in rich, interactive dashboards that can be shared.

Native cross filtering, drilldown, and LLM-enhanced filtering.

Rich Query Editing

All the modern IDE features you expect, built-in autocomplete, type checking.

MCP Server

Provides a basic integration to be used with assistants like Claude desktop. Supports executing queries as well as parsing.

Run MCP Locally

Claude Desktop

Trilogy Studio can be run as a local MCP server for Claude. It'll by default only have access to the default public models. Support for custom models coming soon!

Checkout repo, and from root run below:

uv run mcp install pyserver/mcp_server.py

Or directly edit:

On MacOS: ~/Library/Application\Support/Claude/claude_desktop_config.json On Windows: %APPDATA%/Claude/claude_desktop_config.json

"mcpServers": {
    "Trilogy Language Tools": {
      "command": "C:\\Users\\ethan\\.local\\bin\\uv.EXE",
      "args": [
        "run",
        "--with",
        "mcp[cli]",
        "--with",
        "pytrilogy",
        "--with",
        "trilogy-public-models",
        "mcp",
        "run",
        "<repo_root>\\pyserver\\mcp_server.py"
      ]
    }
}

Run Language Server

Tip

Quick Setup: Run pnpm install in the root, followed by pip install -r requirements.txt in the pyserver subfolder. You can then use pnpm run local to start a local instance.

This will run the frontend with Vite as well as the backend language server.

You can confirm in settings that your local UI is resolving to localhost.

There will be a more polished local option in the future.

Deploying As Service

You can build a production copy and serve as a static website. Github Pages is an easy hosting option for frontend and there is an existing actions pipeline to use as a model.

Developing

Contributions loved! See contributing guide for details.

Inspiration

There are lots of good IDEs out here. Trilogy Studio is probably only best if you want to use Trilogy. Some sources of inspiration:

• Dbeaver
• SQL Server Management Studio
• BQ Cloud Console
• Beekeeper Studio
• QuackDB
• SQL Workbench

Tech Stack

Exists thanks to the following:

Frontend

• Vue
• Vite
• Vega/Altair
• Tabulator