Graphene is a data analytics framework built for agents.
Ask questions and build visualizations 10x faster when agents do the work.
In the future, we believe most low-level data analytics work will be done by agents, allowing humans to focus on insights and decision-making. However, today's tools weren't built with agents in mind:
- They are GUI centric. Lots of actions can only be done via GUI and aren't accessible to external agents like Claude.
- They focus on raising the floor at the expense of lowering the ceiling (limited viz types, simplified querying APIs).
- They assume the human user has the tribal knowledge and business context necessary for analysis.
If we really want agents to be more productive with data, an entirely new toolset is needed.
Graphene is:
- Built for agents. Everything is code, written only in languages that are prevalent in training data (SQL, Markdown, HTML). All actions are CLIs; nothing is trapped in a button.
- High-ceiling. Agents can create any visualization that's supported by ECharts, one of the most feature-complete visualization libraries. And Graphene's query language is as powerful as ANSI SQL, which supports 170+ functions, CTEs, subqueries, set operations, window functions, arrays, and more.
- Optimized for agent context. Graphene's SQL language contains a semantic layer which allows metrics and join relationships to be invoked in queries. When combined with agent skills for general business context and best practices, agents perform at human levels of competency.
Importantly, Graphene is open. You can use this project for internal purposes for free, forever, and aren't locked in to a contract with us. More details below.
Graphene pages support visualizations, input components for filtering and dynamic behaviors, and layout modes for monitoring-oriented dashboards vs. narrative-oriented notebooks.
When you deconstruct data analytics into code, CLIs, and coding agents, things that used to be hard become easy:
- Promote metrics from pages into the model, or demote metrics out of the model back into pages
- Bulk refactors in a single atomic commit/PR
- Ability to use extensive skill/MCP ecosystem to augment agent behavior
- Iterate on a dashboard (edit, run, view) without needing to push up to some API or open a SaaS portal
- Validate SQL and page syntax instantaneously as you type
- Set up a recurring agent that de-bloats, consolidates your model over time
Graphene currently supports Snowflake, BigQuery, ClickHouse, and local data (via DuckDB) as data sources. It is easy for us to add more - just ask.
Once your project is set up, simply start the dev server via npm exec graphene serve (or pnpm graphene serve, etc. based on your package manager) and then prompt your coding agent to do analytics work: answer a data question, build a dashboard, edit the model, etc.
A Graphene project can either be a standalone repo or a directory within a larger codebase (such as dbt). It is comprised of:
- Semantic models, via .gsql files. GSQL is both a modeling language and a query language, in the same way that SQL has both DDL and DML.
- Pages, via .md files. Pages are typically used for dashboards, but can also contain notebook-style narratives, documentation, and other visual content.
Graphene itself is a CLI which can be installed via npm (or pnpm, yarn, etc.). The CLI can run and compile GSQL queries, render pages in the browser, check syntax, print screenshots, and more.
Semantic models are defined like so:
table orders (
id BIGINT
user_id BIGINT
amount FLOAT
status STRING
join one users on user_id = users.id -- many orders per user
is_complete: status = 'Complete' -- dimension (scalar expression)
revenue: sum(amount) -- measure (agg expression)
aov: revenue / count(*) -- measures can compose
)
table users (
id BIGINT
name VARCHAR
join many orders on id = orders.user_id
)Models are then queried via select, either directly via CLI or inside a Graphene markdown page like this.
```sql top_customers
select
users.name as name, -- Use the dot operator to traverse the modeled join relationship
revenue -- Invokes the measure
from orders -- A join statement here is not needed
group by 1
order by 2 desc
limit 10
```
<BigValue data="orders" value="revenue" />
<BarChart data="top_customers" x="name" y="revenue" />Graphene's entire documentation ships as an agent skill in the Graphene npm package. The source files are available here.