Home · Quick start · charter.yaml · Sources · Agent · CLI · MCP · About · FAQ

The idea

Data teams already describe their data. They write data contracts: which sources exist, which tables and columns they hold, which fields are sensitive. That description usually sits in a catalog or a spec file and does little for the person who just wants to look at the data.

DataCharter turns that description into a working tool. Your charter.yaml is the catalog: point it at your sources, and you get one local window that can query, join, chart, and profile across all of them. The contract you already maintain becomes the thing you explore through.

What it is

DataCharter is a single local application: a DuckDB federation engine, a FastAPI server, and a web UI, shipped as one Python package. Run one command and you have a workspace in your browser.

• Contract-governed. Sources are declared in charter.yaml, an ODCS-compatible YAML contract. PII columns — declared, or auto-detected — are masked from the agent by default, and you choose exactly what any agent may see, per source, table, or column. Secrets are never in the file; they are ${NAME} references.
• Local-first. One process on your machine, binding to localhost by default. No cloud dependency, no account, and zero telemetry of any kind.
• Federated. Postgres, MySQL, SQLite, BigQuery, SQL Server, files, Iceberg, Delta, and Snowflake, joined through one engine, with filters and projections pushed down to each source.
• Portable by contract. A workspace is one directory: contract, queries, and an example env file travel as a repo. Secrets and local state never do.

Why local, why now

Exploring data should not require standing up a service, granting a SaaS access to your warehouse, or shipping rows to someone else’s cloud. Modern analytical engines run comfortably on a laptop, and most contracts a team writes are small text files. Putting the two together yields a tool that is fast, private, and easy to share as source.

Keeping it local also makes the natural-language agent honest. The agent is optional and can run entirely on your machine against a local model, on your Claude Code subscription, or on any hosted endpoint. Whichever you pick, PII is masked before anything is sent (and you control what else the agent may see), and the engine stays read-only regardless of what the model suggests.

What it is not

DataCharter is deliberately small. It is not a BI platform, not a data catalog service, not an orchestration tool, and not a multi-user application with its own auth system. It is a local tool for exploring the data your contracts already describe. Its extension surface is DuckDB’s own extension ecosystem rather than a bespoke plugin system.

Built on open source

DataCharter stands on excellent open-source foundations: DuckDB for the engine and federation, the Open Data Contract Standard for the contract format, Vega-Lite for charting, and Monaco for the SQL editor. It is released under the Apache-2.0 license, which permits commercial use.

DuckDB is a trademark of the DuckDB Foundation. DataCharter is an independent project and is not affiliated with or endorsed by the DuckDB Foundation.