Loop
View docs
How it worksThe loopWhy LoopPricingDocs
compare Loop vs OpenStreetMap

Loop is built on OSM.
Then adds what agents need.

OpenStreetMap is the world’s best open map database — and it’s Loop’s primary data source. But OSM alone isn’t agent-native: Overpass QL is not a search API, there are no confidence scores, no freshness labels, and no way for an agent to report what actually happened at a location. Loop seeds from OSM, cross-confirms with Foursquare OS Places, adds real-time verify(), and wraps it all in an MCP endpoint.

Start building →All comparisons
feature comparison
FeatureLoopOpenStreetMap (Overpass API)
Product typeReal-time query API (MCP + REST)Community map database (Overpass API + planet dumps)
Agent accessMCP endpoint — agents connect with one URLOverpass QL — agents must write and parse custom queries
Query interfaceNatural language search + typed filtersOverpass QL query language (e.g. node[amenity=restaurant])
Freshness labelsobserved_at + confidence score on every recordLast-edit timestamp (editor-maintained, not observation-based)
Live verifyverify() re-checks a specific record against OSM liveNo verify concept — data is current when someone edited it
Feedback loopreport() mutates record confidence and freshnessNone — agents can't feed outcomes back to map quality
Cross-source confirmationOSM + Foursquare OS Places (312 of 424 confirmed)Single community source — no cross-confirmation
Typed restaurant schemacuisine[], price_band, outdoor_seating, vegan, vegetarianFree-form tags (cuisine=italian, outdoor_seating=yes) — parse yourself
Availability labelsinferred: true until verified — always labeled honestlyopening_hours tag (when present) — no inferred vs confirmed distinction
CoverageRestaurants & salons, Kreuzberg Berlin (expanding by demand signals)Global — all categories, all countries
Infrastructure requiredNone — call the MCP or REST APIOverpass instance or planet dump + tag parsing pipeline
LicensingODbL (OSM-derived records) + Apache 2.0 (Foursquare OS Places)ODbL — same license, Loop inherits it for OSM-derived data
Out-of-coverage responseExplicit honest error with coverage hintNo concept — Overpass returns whatever matches your query
honest framing

OSM wins on coverage. Loop wins on agent-readiness.

OpenStreetMap has global coverage across every category — roads, transit, parks, buildings, POIs of every type. Loop is Kreuzberg restaurants today. If your agent needs a bus stop in Bangkok or a hiking trail in the Alps, OSM is the answer.

But if your agent is making decisions in a restaurant search pipeline — routing users, booking reservations, answering “is it open tonight?” — Loop gives you what OSM cannot: a pre-parsed typed schema, a confidence score, a freshness timestamp from a real observation, and a verify() tool that re-checks the specific record live before your agent commits.

querying both approaches
OpenStreetMap via Overpass
// Overpass QL — write and parse yourself
[out:json];
node
  [amenity=restaurant]
  [cuisine=italian]
  (52.48,13.39,52.50,13.42);
out body;

// You get raw OSM tags:
// { "tags": { "name": "...", "cuisine": "italian",
//   "outdoor_seating": "yes",
//   "opening_hours": "Mo-Su 12:00-23:00" } }
// No confidence. No observed_at. No verify().
Loop via MCP — same data, agent-native
// Natural language, typed schema
{
  "name": "search",
  "arguments": {
    "query": "Italian restaurant, outdoor seating",
    "location": "Kreuzberg, Berlin",
    "filters": { "outdoor_seating": true }
  }
}

// You get typed JSON:
// { "name": "...", "cuisine": ["italian"],
//   "outdoor_seating": true,
//   "observed_at": "2026-06-12T...",
//   "confidence": 0.91,
//   "availability": { "inferred": true } }
frequently asked questions
When should I use Loop instead of querying OSM directly?
Use Loop when you need to query local business data at runtime inside an AI agent pipeline. Loop gives you natural-language search, typed restaurant schema, freshness signals, real-time re-verification, and a report() feedback channel — no Overpass QL, no tag parsing, no infrastructure. OSM via Overpass is the right choice when you need global map coverage, non-restaurant data, or are building your own data pipeline.
When should I query OSM directly instead of Loop?
Query OSM directly when you need global coverage, non-restaurant POI types (parks, bus stops, buildings), or need to run complex spatial queries (bounding boxes, route proximity). Also appropriate if you're building a data pipeline and want to control the raw data layer yourself. Loop is not a replacement for OSM — it's a runtime service built on top of it.
Does Loop credit OpenStreetMap?
Yes. Loop's restaurant records are seeded from OSM (Overpass API) and Loop uses the ODbL license for those records. verify() re-checks live against OSM on demand — the source field on observations will say 'osm_recheck'. Loop is a consumer and supporter of the OSM ecosystem.

Add Loop to your agent.

OSM is Loop’s source. Loop is what your agent calls at runtime. Add https://stayinloop.dev/mcp — four tools, free tier, no credit card.

Read the docs →Claude setup