Start wide
Explore is good for discovering which pack, place, or pattern matters. The chat can route across multiple maintained datasets without asking you to build a corpus first.
Ask a geographic question and see the answer on the map. Explore is the fastest path from curiosity to maintained geographic context.
Start with one question or open one ready-made map state. Either way, the goal is the same: one visible angle, one obvious next question.
Example queriesOpen Explore with the earthquakes pack ready so the first question is immediately geographic and visible.
Open in Explore › Local climateStart from a concrete county-scale climate story where the map itself makes the correlation legible fast.
Open in Explore › EconomicUse the currency pack as a fast compare surface for questions like who strengthened, who weakened, and where to inspect next.
Open in Explore ›Upcoming climate view for anomaly maps, warming patterns, and broad planetary context.
Explore is the broad map-first lane. Ask across the published packs, let the runtime route the question, and use the map to orient yourself before committing to slower Research or fresher Ops work.
Explore is good for discovering which pack, place, or pattern matters. The chat can route across multiple maintained datasets without asking you to build a corpus first.
Use Explore when the question spans larger historical ranges, multiple regions, or comparative map views that go beyond the live Ops window.
Switch to Research when you want a bounded, repeatable corpus and a longer analytical thread. Switch to Ops when the question is about what is happening now and over the last few days.
Ask the question, watch the result render, and click any feature to see the row behind it. The point is orientation, comparison, and the next question, not GIS setup.
No. Explore opens on maintained packs that are already loaded and joined on loc_id. There are no layers to configure, no projections to pick, and no files to import before the first question.
Every rendered feature comes from a row in a published pack. Click any feature to see the underlying record and the source agency behind it, so a map result stays traceable to its origin.
Resolve your rows to loc_id first, then they join against the same maintained packs Explore reads. Test a coordinate sample, then use bounded MCP tools for programmatic lookups.
The hosted app is free to open and free packs return rows directly. Paid packs and larger agent or bulk workloads are priced separately, and the engine is open source if you would rather run it yourself.