Misata generates referentially consistent multi-table datasets from plain English prompts, YAML schemas, or existing database introspection. The MCP server exposes five tools to Claude and other clients: generate from natural language stories, preview inferred schemas before committing, validate existing schemas, introspect live database connections, and produce Oracle reports that verify row counts, foreign key integrity, and constraint satisfaction. You can ask for "a Brazilian fintech dataset with 1,000 customers and 3% fraud" and get back CSV paths plus table previews. Built for seeding dev databases, generating integration test fixtures, and prototyping BI dashboards without touching production data. Optional LLM enhancement layer supports richer semantic inference across Groq, Anthropic, and local Ollama models.
claude mcp add --transport stdio rasinmuhammed-misata uvx misata