Graphene: Data Analytics Framework Built for Coding Agents

Agent-Centric Analytics Framework

Graphene is a data analytics framework explicitly designed for coding agents as the primary user persona. It aims to replace traditional BI tools by leveraging the context and capabilities of LLMs to perform end-to-end data tasks, from answering questions to building dashboards.

The framework addresses two critical gaps for agents in data work:

• Semantic Layer: Uses Graphene SQL to combine standard SQL power with semantic governance (measures, modeled joins), yielding more accurate queries.
• Dashboard File Type: Provides a structured format for visuals that is more consistent and polished than raw Python or JavaScript outputs.

Technical Architecture and Features

Graphene operates as a CLI tool installable via npm, allowing agents to compile queries, render pages, and check syntax. It supports Snowflake, BigQuery, ClickHouse, Postgres, MotherDuck, and local data via DuckDB.

Key technical components include:

• Graphene SQL: Inspired by Malloy, it supports CTEs, subqueries, window functions, and over 170 functions. It allows agents to define semantic models (tables, joins, measures) in .gsql files.
• Graphene Markdown: Pages are defined in .md files, mixing SQL queries with visualization components (e.g., BigValue, BarChart) and standard HTML/CSS/JS for custom layouts.
• Agent Skills: The entire documentation ships as an agent skill within the npm package, enabling agents to self-learn and iterate without external context.

Design Philosophy and Licensing

The project emphasizes token efficiency and agent ergonomics, minimizing boilerplate to reduce context window usage. It promotes an "everything-as-code" approach, allowing version control, CI testing, and easy refactoring of analytics logic alongside application code.

Graphene is licensed under the Elastic License 2.0, permitting free internal use forever. The business model includes Graphene Cloud, a hosted solution for dashboards, an MCP server, and a Slack bot for commercial use.