Query all your data locally — then hand AI agents exactly the data you choose, and not one column more.
One process, your machine, no cloud dependency. Zero telemetry. Connect an agent — Claude Code, a local model, or any OpenAI-compatible endpoint.
No global install, no config, nothing to sign up for. If you have uv, you have a workspace.
# Try it instantly on generated demo data: uvx datacharter serve # → opens http://127.0.0.1:8321 # Or install it: pip install datacharter # Start a real workspace: datacharter init # scaffolds charter.yaml, queries/, .env.example datacharter serve # explore in your browser
uvx datacharter serve launches a local workspace on generated demo data. The engine, API, and web UI all run in one process.
Browse the source tree, write SQL in the editor, chart and profile results, and drag a CSV, Parquet, or JSON file onto the window to query it instantly.
datacharter init scaffolds a workspace: charter.yaml + queries/*.sql + .env.example. Commit it, clone it, serve it. The whole environment travels as a repo.
A single local tool for querying, charting, and understanding every source your data team already describes.
charter.yaml describes sources, tables, and PII fields: the same ODCS-compatible contract spec your data team already writes.
Postgres, MySQL, SQLite, BigQuery, SQL Server, files, Iceberg, Delta, and Snowflake. Federated joins across all of them, with filters and projections pushed down to each source.
One process on your machine, no cloud dependency, no phone-home of any kind. Binds to localhost by default; the engine is read-only.
Contract, queries, and example env travel together as a repo. Secrets and local state never do; git clone && datacharter serve reproduces the whole environment.
A Monaco SQL editor with catalog autocomplete, a virtualized results grid, Vega-Lite charts, a profiling panel, an EXPLAIN viewer, and CSV/Parquet/JSON/XLSX export.
Ask in plain language via your Claude Code subscription, a local Ollama model, or any OpenAI-compatible endpoint — then choose exactly what it sees, per source, table, or column. PII is auto-detected and masked by default; flip Agent view to confirm what it gets back.
Source tree, SQL editor, results, charts, profiling, and an optional chat rail: one local window over all your data.
DataCharter stands on excellent open-source foundations.
DuckDB is a trademark of the DuckDB Foundation. DataCharter is an independent project and is not affiliated with or endorsed by the DuckDB Foundation.