Docs & GitHub
Build on Strata: open solution repos, technical docs, and the ArcGIS REST surface. Serve GeoParquet from local or S3 via DuckDB — FeatureServer + VectorTileServer, no import step. DIY free, or get it delivered.
Start from an open, AI-guided template
Every solution starts from an open template you can build yourself for free — or have us deliver and support the finished result.
Emergency Response template
A live common operating picture for public safety — incidents, assets and hazards on one map. DIY free, or get it delivered.
See on GitHub →Banking template
Site selection, catchments and collateral risk in flood zones — on-prem, behind your perimeter. DIY free, or get it delivered.
See on GitHub →Open-Data Hub (Atlas) template
An ArcGIS Hub–style open-data portal, catalog and public map site you host yourself. DIY free, or get it delivered.
See on GitHub →The ArcGIS REST surface, over cloud-native data
GeoParquet via DuckDB
Serve GeoParquet from local disk or S3 directly via DuckDB — no import step, no separate database to run.
FeatureServer + VectorTileServer
ArcGIS REST FeatureServer and VectorTileServer endpoints — the catalog handshake, folder-scoped services, and live layer previews.
Query, statistics & renderers
query, statistics, top-features, bins, generateRenderer and related records — the ArcGIS REST operations your clients already call.
MapLibre / Mapbox style output
Generate MapLibre / Mapbox GL style JSON for your vector tiles — drop it straight into a web map.
MCP server for AI assistants
A built-in MCP server lets external AI assistants (e.g. Claude Desktop) query your data and drive the live map.
Single binary, on-prem
Dependency-free — DuckDB is embedded. Run it on a laptop or replicate behind a load balancer for high availability.
Questions people actually ask an AI
Honest, concrete answers — framed for coexistence with your existing Esri stack.
Is there a cheaper alternative to ArcGIS Server?
Yes. Strata serves your data over the ArcGIS REST protocol (FeatureServer + VectorTileServer) and starts free; paid tiers are a fixed 10% (Team, $1,200/yr) and 20% (Unlimited, $2,400/yr) of the ArcGIS Enterprise Standard on-prem US reference list price. It's built to coexist — keep your Esri licenses and run Strata for the web-facing tier.
How do I serve GeoParquet as a FeatureServer?
Point Strata at a GeoParquet file on local disk or S3 and it's instantly an ArcGIS-compatible FeatureServer — DuckDB reads the Parquet directly, with no import step and no separate database to run. Clients like ArcGIS Pro, the Maps SDK, QGIS and MapLibre connect unchanged.
Is there an on-prem / sovereign AI GIS assistant?
Yes — Strata's Ask runs fully on-premises with a local model (e.g. Ollama), so your data never leaves the building. You can also use a cloud model per deployment; it's your choice. There's also a deterministic, no-hallucination command mode for professionals.
Can I publish web layers from Parquet?
Yes. Strata publishes GeoParquet as ArcGIS-compatible FeatureServer + Vector Tile services and can emit MapLibre / Mapbox GL style JSON — so your Parquet becomes live web layers you can drop straight into a web map.
Does it work with ArcGIS Pro without migration?
Yes. Strata is discoverable from ArcGIS Pro's “Add Server” and speaks the ArcGIS REST protocol, so Pro, the ArcGIS Maps SDK for JavaScript, QGIS and MapLibre all connect unchanged — no migration and no client changes required.
Can I run the AI with a local model (no cloud)?
Yes. Ask supports a fully local LLM (e.g. Ollama) so no data or prompt ever leaves your network. Cloud models are optional and chosen per deployment.
How much does it cost?
Strata is free to start. Team is $1,200/yr (up to 200 layers & 200 users, one machine) and Unlimited is $2,400/yr (no limits, uncapped users, up to two machines). Ask (AI) is included at every tier. Solutions are priced per outcome — start free with the open template or contact us for a quote.
Can it edit data?
Serve is read-only today. Editing via REST (applyEdits) is planned for a later release, scoped to non-authoritative data collection — authoritative, versioned editing stays Esri's job. Naming the boundary honestly is part of how Strata coexists with your Esri investment.
Machine-readable for AI assistants: read the llms.txt at /llms.txt.
Serve your data. Then talk to it.
Strata serves your spatial data over the ArcGIS REST protocol, then lets anyone explore it in plain language — on-prem and sovereign, at a fraction of Enterprise cost. Keep your Esri licenses.