Show HN: Dinobase — Agent-First Database Providing SQL Across 100+ Sources with 91% Accuracy vs 35% Baseline

Hacker News·low signal

Dinobase launched as an agent-first database where each SaaS API, database, and file storage becomes a SQL schema that AI agents can query and mutate. In benchmarks across 11 LLMs, agents achieved 91% query accuracy vs 35% with raw APIs, 3x faster, and 16-22x cheaper per correct answer. Provides native SDKs for LangChain, CrewAI, LlamaIndex, and Pydantic AI, plus an MCP server for Claude Desktop and Cursor. Addresses a real pain point: giving agents reliable data access without per-source integration work.