Skip to main content

​
What is Datagate?

Datagate is a data marketplace that connects AI agents to premium datasets. Sellers publish vector-indexed datasets, and buyers query across them via a single API call. For AI agent builders, Datagate provides:
  • One API to search across multiple sellers’ datasets simultaneously
  • Automatic embedding — send natural language, the server handles vectorization
  • Cross-model search — query datasets using different embedding models in a single call
  • Metadata filtering — filter results per-dataset using MongoDB-style operators
  • Pay-per-chunk pricing — only pay for the results you receive

​
How it works

Your Agent → POST /v1/query → Datagate → Fan-out to seller vector DBs → Merge & rank → Results
  1. Discover datasets with list_datasets() — see what’s available, pricing, and filterable fields
  2. Query with query() — send natural language text, get ranked results with metadata
  3. Pay per chunk — each result chunk is billed at the dataset’s price (typically fractions of a cent)

​
Two ways to integrate

SDK — use Datagate directly in your code. Returns typed objects, full control over error handling. 
pip install datagate     # Python
npm install datagate     # TypeScript
MCP server — connects Datagate to Claude Desktop, Claude Code, or any MCP-compatible agent. No code needed — the AI agent discovers and calls the tools automatically. The MCP server uses the SDK under the hood. 
pip install datagate-mcp     # Python
npm install @datagate/mcp    # TypeScript

​
Get started

Quickstart

First query in under 5 minutes

Python SDK

pip install datagate

TypeScript SDK

npm install datagate

MCP Integration

Connect to Claude Desktop or Claude Code