EmbedEarth (E2) MCP server

EmbedEarth

Open, programmable geography for applications and AI agents.

EmbedEarth (E2) is an open, programmable geographic engine for the physical world. Search places and real-world features, run spatial computations, build routes, render maps, and work with geographic datasets through one SDK.

E2 is built as an open alternative to Google Maps Platform, bringing together capabilities commonly spread across Maps, Places, Routes, Geocoding, Autocomplete, and Street View — while adding spatial computation, open geographic data, and local execution.

Get started · Documentation · Examples

EmbedEarth is an early developer foundation. The API surface, hosted services, and open-source boundary are still being shaped.

Packages

Modes

E2 can run locally or in the cloud. Local datasets are downloaded once per region and reused for future requests, so the same data does not need to be fetched again.

Why E2

Traditional mapping APIs are built primarily around maps, places, and directions. E2 makes the broader physical world programmable.

• Search places, infrastructure, observations, nature, civic data, and other real-world features.
• Compute distance, proximity, counts, density, coverage, gaps, and other spatial relationships.
• Map basemaps, features, routes, and application data.
• Route between coordinates or directly to real-world features.
• Stratum geographic datasets for local, offline, and cloud workflows.
• Access street-level imagery and visual observations tied to geographic locations.

For developers coming from Google Maps Platform, E2 can replace many common workflows built around Maps, Places, Routes, Geocoding, Autocomplete, and street-level imagery while giving applications direct access to broader geographic datasets and spatial computation.

Try it

The repository includes a TypeScript SDK, CLI, MCP server, Stratum local-data workflows, and spatial compute primitives.

pnpm install
pnpm build
pnpm --filter @embedearth/cli exec earth search restaurant --area Brooklyn

From TypeScript:

import { E2 } from "@embedearth/sdk";

const earth = new E2({ apiKey: process.env.E2_API_KEY });
const results = await earth.search({
  feature: ["restaurant"],
  area: ["Brooklyn"],
  mode: "auto",
});

Six core areas

Search

Search the physical world by feature, place, region, or coordinates. Query places, infrastructure, observations, natural features, civic data, and other geographic datasets through one interface. Results are returned as GeoJSON and can be passed directly into Compute, Map, or Route.

await earth.search({ feature: ["restaurant"], area: ["New York"], mode: "auto" });

Compute

Run spatial analysis directly against geographic features. Find what is nearest, measure distance, count features, calculate density and coverage, detect gaps, or cluster nearby objects — turning geographic data into answers without a separate GIS stack.

await earth.compute.feature("restaurant").area("New York").near("park", 500).count();

Map

Render E2 data using MapLibre-compatible maps and PMTiles. Load the global basemap alongside feature layers, Search results, routes, and your own geographic data, while your application keeps control over styling and interaction.

Shortened. The full README is on GitHub.