DATA FOR AI AGENTS · HISTORICAL GEOGRAPHY

Verified historical geography for your AI agents.

AtlasPI is a REST API + MCP server for historical geography: structured, cited data that cuts LLM hallucinations on historical queries by 3×. Instead of making things up, the agent retrieves facts with sources. Open source, with a citable Zenodo DOI.

Stack REST API MCP Server 1,038 entities 4500 BC–2024 Apache 2.0

What it does

AtlasPI gives AI agents structured, verified historical-geography data — entities, cities, events, trade routes — with confidence scores and academic sources. Instead of inventing, the agent retrieves cited facts via REST API or MCP server, measurably reducing hallucinations on historical queries.

Who it's for

  • Developers of AI agents and RAG applications
  • Digital-humanities and research projects
  • EdTech with interactive maps and timelines
  • Game design that needs historical accuracy
Features

Cited historical facts, not invented ones

REST API

50+ endpoints, structured JSON, no authentication to get started.

MCP server

Native integration with Claude Desktop and Model Context Protocol agents.

Cited data

Academic sources, confidence scores and confirmed / uncertain / disputed status.

5,000 years of coverage

From 4500 BC to 2024, with an interactive map and time slider.

Fewer hallucinations

Measured error-rate drop from 0.21 to 0.07 per query (≈3×).

Open source

Apache 2.0 core, with a citable Zenodo DOI for papers.

1,038
structured entities
4,819
academic sources
−3×
hallucinations on historical queries

Build AI agents that don't invent history

Explore the AtlasPI API and MCP server, or contact us for high-volume or dedicated-data use.