Most of the data your agent needs about a place on Earth is already public. The Census geocoder is free. NWS weather is free. FAA's flight tracker, the IRS exempt-org database, every Socrata-powered city portal. All free.
The problem is they're each a different shape. Some want lat/lon. Others want FIPS codes. The FAA endpoint wants IATA codes, the IRS wants a 9-digit EIN. Half of them return XML if you ask wrong. Rate limits are inconsistent across the board. And none of them speak x402, so an agent that wants to bill its caller a penny per trip-plan can't pass the cost through cleanly.
That's what Locale fixes.
What's in the cluster
Five endpoints right now. All $0.001 per call, settled in USDC on Base:
geocode: address string in, lat/lon plus county FIPS out. Wraps the Census batch geocoder, cached.weather: lat/lon in, current observation plus 7-day forecast out, normalized JSON. NWS upstream.flight-status: IATA flight number plus date in, gate/status/delay-minutes out. FAA SWIM feed.nonprofit-lookup: EIN or org name in, IRS BMF row plus latest 990 PDF link out.civic-dataset: Socrata dataset ID plus query in, normalized rows out. Works against any city portal that exposes Socrata.
Each one is a thin wrapper. No model in the loop, no rewriting, no opinion about what the answer should look like. The shape is whatever the upstream gives you, with field names lowercased and units made consistent.
A trip in eight calls
Say a travel agent gets asked: "Book me a Friday day trip from SFO to Seattle for the food and the public art."
Here's what eight calls of Locale buys it, for one cent total:
1. geocode SFO. Confirms the airport's coordinates.
2. flight-status for the user's existing inbound flight Thursday night. Confirms they'll actually be in SF Friday morning.
3. weather Seattle Friday. 58°F and a 30% chance of rain in the afternoon. Plan accordingly.
4. geocode "Pike Place Market, Seattle". Lat/lon for the lunch hub.
5. civic-dataset Seattle's public-art-installations Socrata feed, filtered to a 1km radius around Pike Place. List of murals and sculptures with addresses.
6. civic-dataset Seattle's food-establishment-inspections dataset, filtered to the same area, sorted by recent inspection score. Which restaurants to actually walk into.
7. nonprofit-lookup Seattle Art Museum. Hours and EIN-confirmed legitimacy.
8. flight-status for the user's return flight. Gate and delay window, so the agent knows whether to push dinner earlier.
Cost: $0.008 of USDC, settled through the standard x402 facilitator path on Base. The agent now has an itinerary built from primary sources, no scraping or SerpAPI middleware involved, no model inventing a restaurant that closed in 2022.
You're paying for shape, not data
This is the thing that confuses people the first time. "Why pay anything if NWS is free?"
Because the wrapper isn't free. Consistent JSON, no rate-limit surprises, no XML, no signing dance with FAA, no per-portal Socrata auth, and a single x402 payment path your agent already speaks. You pay a penny per call to skip writing five different clients.
If you were going to call NWS once a week from your laptop, write the curl yourself. But if you're an autonomous agent making 200 trip-plans a day and you don't want to babysit five upstream APIs with five auth schemes, Locale costs $0.20 of USDC and an afternoon you don't have to lose.
What's coming
A few more datasets are queued: USGS earthquake feed (useful for travel and insurance agents), TSA wait-times, NOAA tide-and-current feed for coastal logistics, and a unified court-records search across the handful of state portals that don't charge for access. All at $0.001, all under the Locale cluster.
The full list is at agentutility.ai/clusters/locale. agentId 47167 on Base if you want to verify the cluster on-chain before you wire a call.