Headless CMS for geospatial content

The spatial backend built for the AI era.

A fully managed, AI-native platform for geospatial content. Your editors manage map data without code. Your developers deliver it through fast, cacheable APIs. AI automates the work in between.

Works with the map tools you already use

  • Esri
  • MapLibre
  • Mapbox
  • MapTiler
  • QGIS
One platform, three jobs

Editors, developers and AI agents work from the same source of truth.

For editors

Manage places and content without code

Add points, lines and areas on the map, then enrich them with text, images and files. Draft changes safely, then publish when ready.

For developers

Build map apps without a backend

A stable API, GeoJSON and vector tiles serve your published content straight to your apps. Nothing to run, and the data is always up to date.

For AI agents

Let AI do the tedious work

Connect Claude, ChatGPT or your own agent through MCP to answer questions, import, fill gaps and clean up. Drafts by default, live only when you allow it.

[ image ] Screenshot: Type builder, a "Trailhead" type with its fields (name, difficulty, photos, location) and the type tree on the left
How it works

From raw data to live maps in four steps.

Start and model the data once: the kinds of places and the fields each one has. Editors add content by importing it or drawing on the map, then publish it for your apps to consume through a cacheable API and vector tiles. Put AI agents to work to answer questions, import and enrich your data, as drafts for editors to review or live when you allow it.

AI

Hours of map admin, done while you review.

Connect the AI assistant you already use. It works through your own types and fields: imports spreadsheets, writes missing descriptions, cleans up duplicates and answers questions about your places. You decide what it may publish.

Activity · AI assistantsTODAY
✎Descriptions draftedTrails · Content bot23
⌖Imported from a spreadsheetBike shops, placed from addresses · Ana via Claude84
⚑Duplicates moved to trashRestorable for 30 days6
↑Published with permissionMountain huts · Ops bot31
✓Waiting for reviewDrafts changed by AI assistants107
Ask your datavia MCP
Which campsites within 2 km of Lake Tahoe have no description?
Found 12 campsites without a description. I drafted one for each from its facilities, season and nearby trails.
12 drafts0 publishedReview changes →
Delivery API

Published content, ready to consume.

Cached, ETag-aware and CORS-enabled. Use JSON, GeoJSON or Mapbox vector tiles with the map library you already use.

  • JSON, GeoJSON and vector tiles
  • Draft and published are separate
  • Works with MapLibre, Mapbox, QGIS, ArcGIS
JSONGeoJSONVector tiles
GET/delivery/v1/projects/{id}/entries.geojson
200 OKETagcache HIT18 ms
{
  "type": "FeatureCollection",
  "features": [{
    "type": "Feature",
    "geometry": { "type": "Point",
      "coordinates": [-122.41, 37.77] },
    "properties": {
      "name": "Ocean Beach",
      "type": "beach"
    }
  }]
}
FAQ

Questions, answered.

What is Spatibase?

Spatibase is a headless CMS for geospatial content. Editors manage points, lines and areas on a map without code, and developers get a cacheable API that serves published content as JSON, GeoJSON and vector tiles, so they can build map apps without running a backend.

How is it different from a general headless CMS such as Strapi or Contentful?

Locations are first-class. Entries have real geometries, editors work on a map, and the platform publishes GeoJSON and vector tiles out of the box, instead of storing coordinates as plain text fields.

How is it different from a GIS or a map platform such as Mapbox or ArcGIS?

Spatibase manages the content behind your map: the places, their fields, media and the editorial workflow. You keep using the map library or GIS you already have, such as MapLibre, Mapbox GL or QGIS, to render it.

Do I need to build a backend for my map app?

No. Published content is served directly by the Delivery API, which is cacheable and always up to date, so your app reads the data without a server of its own.

Can AI assistants work with my data?

Yes, through MCP: connect Claude, ChatGPT, Cursor or your own agent with a key or a sign-in you approve. A read key answers questions about published content; an edit key creates and changes drafts for an editor to review. Publishing and deleting are extra switches per key, off by default, and every change is in the activity log.

Which formats can I import and deliver?

Import GeoJSON, CSV and TSV, and upload GPX and KML shapes. Deliver JSON, GeoJSON and Mapbox Vector Tiles with TileJSON.

Enterprise

Built for organizations.

Single sign-on, roles, audit logging and support to match how your team works. Run it managed, on dedicated infrastructure, or self-hosted in your own cloud.

Illustration of a rack-mounted server

Build your spatial backend

Start free, or talk to us about your organization.