io.github.datacharter/datacharter MCP Server
io.github.datacharter/datacharter is a local-first data explorer for AI agents that provides read-only, PII-masked query tools. It exposes a local SQL workspace over files and databases, federated by DuckDB, and governed by a charter.yaml contract that defines row-level security and granted access.
🛠️ Key Features
- Local SQL workspace over local files and databases
- DuckDB federation for querying across data sources
charter.yamldata-contract governance- Read-only query tools
- PII masking for agent-accessible results
🚀 Use Cases
- Governed data exploration by AI agents
- Enforcing contract-defined access using
charter.yaml - Querying without granting “one column more” than the contract allows
⚡ Developer Benefits
- Aligns agent tooling with model context via MCP
- Uses SQL as the query interface
- Supports self-hosted deployment
- Contract-driven data governance with row-level security
⚠️ Limitations
- Provides access as defined by the
charter.yamlcontract - Read-only querying only
Supporting Topics
data-contracts, data-engineering, data-exploration, duckdb, local-first, mcp, python, sql, data-catalog, data-governance, mcp-server, model-context-protocol, pii-masking, row-level-security, self-hosted, odcs, pii
