datacharter 0.10.0

Charter your data — contract-governed local data exploration, powered by DuckDB

DataCharter

Query all your data locally — then hand your AI agents exactly the data you choose, and not one column more.

The big-words version: a local, federated data explorer with governed, regulated agentic data access, powered by DuckDB. Here's what that actually means 👇

🔍 Query all your data, locally — no pipelines, no warehouse, no waiting

• Local CSV, Parquet, and JSON files
• Postgres, MySQL, SQL Server, Snowflake, BigQuery — and more
• JOIN a local CSV → a Snowflake table → a Parquet file in S3, in one SQL statement, all on your laptop
• Yes, it's as unreasonable as it sounds. You kind of have to try it to believe it.

🤖 Connect an agent — and decide exactly what it's allowed to see

• Claude Code — runs on your existing subscription, no API key
• A model running fully local with Ollama
• Any OpenAI-compatible agent
• Grant or deny access in the UI or right in your data contracts, at every level: whole sources → individual tables → individual columns
• PII is auto-detected and defaulted to no agent access — override per field if you really mean to
• Don't take our word for it: flip on Agent view and see, column by column, exactly what your agent gets back when it runs a query. (Spoiler: the PII comes back •••.)

Wait, there's more!

Beyond local federation and governed agent access, you also get:

• See answers as you type. Live results preview while you write SQL, one-click auto-charts, and a profiling panel (missing values, distributions, outliers) — no separate BI tool.
• Safe by design. The engine is read-only by construction — no query can write, delete, or touch the filesystem — so pointing an AI (or a teammate) at your real databases can't do damage.
• Point other AI tools at your data, too. A governed MCP server exposes the same read-only, PII-masked query tools to Cursor, Cline, or your own agent.
• Trust every answer. Each result shows exactly which source columns it read — so you always know where a number came from.
• Save, reuse, export. Snapshot a result as a reusable local table; export to CSV, Parquet, JSON, or XLSX.
• Governance you can automate. Catch schema/PII drift in CI, diff data across sources, and define certified metrics — all from the command line.

Status: pre-release. V1 in development.

Quick start

# Try it instantly on generated demo data — no install, no config:
uvx datacharter serve          # needs `uv` → https://astral.sh/uv
# → serves at http://127.0.0.1:8321 (open it in your browser)

# Or install it:
pip install datacharter        # Python 3.11+

# Start your own workspace:
datacharter init               # scaffolds charter.yaml, queries/, .env.example
# → add a source: edit charter.yaml, or use the "Sources" panel in the UI
datacharter serve              # → http://127.0.0.1:8321

Then, once it's running, drag a CSV, Parquet, or JSON file onto the window to query it instantly — no config needed.

Optional natural-language agent — point it at any OpenAI-compatible endpoint:

export OPENAI_BASE_URL=...     # any OpenAI-compatible API
export OPENAI_API_KEY=...
datacharter serve

…or run fully local — no API key, no data leaves your machine (requires Ollama):

ollama pull qwen3:8b           # once
datacharter serve --local      # qwen3:8b by default (--model to change)

Why DataCharter

• Your contracts are the catalog. charter.yaml describes sources, tables, and PII fields — the same contract spec your data team already writes, so there's no separate metadata store to maintain.
• Real federation, not just a shared connection. Filters and projections are pushed down to each source — even across a cross-source join, every leg is filtered where its data lives. (Snowflake runs via connector extract, datacharter[snowflake], with the same pushdown into the extract.)
• Local-first. One process, your machine, no cloud dependency. The optional --local agent runs a small open model via Ollama — no API key, no data leaves your machine.
• The workspace is a directory. charter.yaml + queries/*.sql + .env.example — commit it, clone it, datacharter serve. Your team's whole exploration environment travels as a repo; secrets and local state never do.

DataCharter governs and audits your data, not just displays it. The full command set (drift, scan, diff, metric, mcp, and more) is in the CLI reference; the security model is in security.

Built on

DataCharter stands on excellent open-source foundations:

• DuckDB — the analytical engine at our core: federation (ATTACH), file formats, Iceberg/Delta, encryption, autocomplete.
• Open Data Contract Standard / datacontract.com — the contract format charter.yaml speaks.
• Model Context Protocol — the open protocol the datacharter mcp server speaks to agents and MCP clients.
• Vega-Lite — declarative charting.
• Monaco Editor — the SQL editor.
• TanStack Table & Virtual — the virtualized results grid.
• And the Python & React ecosystems — FastAPI, pydantic, httpx, keyring, and ruamel.yaml on the backend; React and Vite on the front.

Testing uses VidaiMock, an Apache-2.0 mock LLM server, as the offline agent endpoint in CI.

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

License

Apache-2.0