# DataCharter

> Contract-governed local data exploration, powered by DuckDB. One local SQL
> workspace over your files and databases (CSV, Parquet, JSON, Excel, Postgres,
> MySQL, SQLite, SQL Server, Snowflake, BigQuery, Iceberg, Delta), with governed,
> read-only, PII-masked access for AI agents via chat, Claude Code, or MCP.
> `pip install datacharter` / `uvx datacharter`. Apache-2.0.

Key facts: the engine is read-only by construction; PII is auto-detected and
masked from agents by default; access is granted per source, table, and column
in a `charter.yaml` data contract; every agent access is logged to a
tamper-evident, hash-chained audit trail; plain-English policies (aggregates
only, k-anonymity thresholds, join limits) are enforced by query analysis;
opt-in canary tripwires alarm if a masked value ever escapes to agent output.

## Docs

- [Quick start](https://datacharter.dev/quickstart.html): install, first workspace, first query
- [The workbench](https://datacharter.dev/editor.html): SQL editor, live results, charts, profiling, command palette
- [Guides](https://datacharter.dev/guides.html): markdown agent context, datacharter suggest, measuring lift
- [charter.yaml](https://datacharter.dev/charter-yaml.html): the data-contract file — sources, PII, agent access
- [Sources](https://datacharter.dev/sources.html): every supported file and database source
- [Agent](https://datacharter.dev/agent.html): connecting Claude Code, Ollama, or any OpenAI-compatible agent
- [MCP server](https://datacharter.dev/mcp.html): governed MCP tools for Cursor, Cline, Claude, and others
- [Policies](https://datacharter.dev/policies.html): aggregates-only, k-anonymity, and join policies
- [Audit](https://datacharter.dev/audit.html): the flight recorder and canary tripwires
- [Evals](https://datacharter.dev/evals.html): scoring agent answers and measuring guide lift
- [CLI](https://datacharter.dev/cli.html): every subcommand
- [Workspace](https://datacharter.dev/workspace.html): workspace layout and state
- [Desktop](https://datacharter.dev/desktop.html): the desktop app (beta)
- [FAQ](https://datacharter.dev/faq.html): common questions

## Optional

- [Local DuckDB UI](https://datacharter.dev/duckdb-ui.html): DataCharter as a DuckDB editor/GUI
- [Hex alternative](https://datacharter.dev/hex-alternative.html): local-first, open-source comparison
- [Governed MCP data access](https://datacharter.dev/mcp-data-governance.html): safe AI access to real data
- [Blog](https://datacharter.dev/blog/): release deep-dives and design notes
- [About](https://datacharter.dev/about.html): who builds DataCharter and why
- [Security](https://datacharter.dev/security.html): security model and reporting
- [GitHub](https://github.com/datacharter/datacharter): source, issues, releases
- [PyPI](https://pypi.org/project/datacharter/): the Python package
