LocalData MCP

A data science plugin for LLM agents. Connects to 13 database types, 20+ file formats, graph and RDF sources, and provides 8 analytical domains through 52 MCP tools.

LocalData MCP gives LLM agents direct access to local and remote data sources through the Model Context Protocol. Beyond data connectivity, it provides a full data science toolkit: statistical analysis, time series forecasting, regression modeling, clustering, business intelligence, geospatial analysis, optimization, and sampling methods — all designed for LLM-driven workflows with composition metadata and progressive disclosure.

Key capabilities

  • SQL databases: PostgreSQL, MySQL, SQLite, DuckDB, Oracle, MS SQL Server

  • NoSQL: MongoDB, Redis, Elasticsearch, InfluxDB, Neo4j, CouchDB

  • File formats: CSV, TSV, JSON, YAML, TOML, XML, Excel, Parquet, Feather, HDF5

  • Structured data: JSON, YAML, TOML trees with full CRUD operations

  • Graphs: DOT, GML, GraphML, Mermaid with path finding and statistics

  • RDF/SPARQL: Turtle, N-Triples files and remote SPARQL endpoints

  • Data science: 8 analytical domains with 12 specialized tools

  • Streaming: Memory-bounded query execution with adaptive chunk sizing

  • Security: Path restrictions, SQL injection prevention, connection limits

Contents