Aleph is an open-source Python tool that downloads real-world geographic data for a region you select. Draw a rectangle or polygon on a map and it grabs everything inside it: buildings, elevation/terrain, HD satellite imagery, Street View photos, and textured 3D models. It can also return any of that for a single point on the fly via a CLI. It was built so LLMs can quickly gather information about the physical world, but you can just use it to keep a piece of the Earth on your laptop.

Does Aleph require API keys?

No. Aleph requires Python 3.11+ and only two dependencies (Pillow and numpy), with no API keys needed. It works by calling undocumented APIs, which is also its main caveat — those endpoints can break or hit rate limits without warning.

Is it legal to use the data Aleph downloads?

It depends on the layer. OpenStreetMap data and geocoding are under the ODbL (commercial use allowed with attribution). AWS Terrain Tiles are open with credits. But Street View, satellite imagery, and 3D meshes are © Google and its providers, and downloading them goes against Google's terms of service. Aleph is not affiliated with Google, and the author is explicit that you are responsible for how you use the tool and its data.

How do you use Aleph?

Install with pip, then run 'alephgeo dashboard -o captures' to open a browser dashboard at 127.0.0.1:8100 where you draw a region, pick sources, preview, and capture. Or use the CLI directly for point lookups, e.g. 'alephgeo streetview --place "Colosseum, Rome" --best-match -o captures' or the same with 'satellite'. Captures save locally and you can explore 3D meshes and terrain relief in the browser.

Aleph: Download a Piece of the Real World — Satellite, Terrain, Buildings, and Street View — for You (or Your LLM)

By Prahlad Menon Published 2026-10-22 3 min read

Here’s a pattern worth watching: tools built first for LLMs that turn out to be delightful for humans too. Aleph, by Belluxx, is exactly that. Its tagline — “Download a piece of the world, for you (or your LLM)” — is literal. Draw a box on a map and it pulls the real world inside it onto your laptop.

What It Does

Aleph does two things:

Bulk capture — draw a rectangle or polygon on a map and it downloads everything inside it: building footprints, elevation/terrain, HD satellite imagery, Street View photos, and textured 3D meshes.
On-the-fly point lookups — ask for any of that data at a specific place via the CLI, no dashboard needed.

alephgeo streetview --place "Colosseum, Rome" --best-match -o captures
alephgeo satellite  --place "Colosseum, Rome" --best-match -o captures

Or launch the browser dashboard (alephgeo dashboard -o captures, served at 127.0.0.1:8100), draw your region, pick sources, check the preview and time estimate, and capture. You can then toggle each layer on or off and explore the 3D mesh or terrain relief interactively.

The setup is refreshingly light: Python 3.11+, just Pillow and numpy, and no API keys.

Why It Was Built: Grounding LLMs in the Physical World

The origin story is the interesting part. Aleph “was born as a way to let LLMs quickly gather information from the physical world.” That’s a meaningful gap — language models are fluent about places they’ve read about but have no direct channel to pull the actual satellite tile, the actual elevation grid, or the actual Street View of a location on demand. Aleph is that channel: a clean, dependency-light way for an agent to fetch ground truth about a coordinate or a region.

It fits a trend we keep covering — infrastructure designed to be driven by AI, like SolidCraft’s MCP-drivable CAD kernel or Pixel Path baking OpenStreetMap into playable scenes. Aleph sits a layer below those: it’s the raw-world data pump. Pair it with something like SUPRA-Tex turning images into relightable 3D assets and you can see a pipeline forming — real-world capture feeding generative 3D.

The Honest Caveats

The README doesn’t hide the catches, which I respect:

Undocumented APIs. Aleph leans on unofficial endpoints, so it “may break or hit rate limits” at any time. This is not a stable production data source.
Licensing is a real boundary. OSM data (ODbL) and AWS Terrain Tiles are fine with attribution. But Street View, satellite imagery, and 3D meshes are © Google, and downloading them goes against Google’s terms. Aleph isn’t affiliated with Google, and the author puts responsibility squarely on the user. Treat the Google-sourced layers as “look, don’t ship.”

Why It Matters

Even with the caveats, Aleph is a genuinely useful primitive. For researchers, makers, game designers, and LLM builders, having a one-command way to pull terrain, buildings, imagery, and 3D meshes for an arbitrary patch of Earth — with no keys and two dependencies — collapses a normally painful geospatial workflow into minutes. Use the open layers (OSM, terrain) for anything you plan to redistribute, and keep the Google-sourced layers for personal exploration.

It’s a small tool with an outsized “whoa” factor: a piece of the real planet, on your laptop, ready to hand to a model.

Repo: github.com/Belluxx/Aleph