This server answers questions about where films, TV series, games, anime and manga are placed in the real world, using the Movie Scene Map (formerly FilmMap) atlas built from Wikidata.
Find filming locations for a title (where_was_it_filmed): give a film/series/game title (or slug) and get every place Wikidata records it as filmed at, with coordinates, plus any countries recorded too coarsely to place (countries_only).
Literal places mostly in India
Invert it — what was filmed at a place (what_was_filmed_here): give a place slug or name and get everything recorded as shot there, most widely covered first.
Search productions (search_productions): accent- and case-insensitive substring search, optionally filtered by film, tv, game, anime or manga — useful when unsure of a title.
Search locations (search_locations): find places by name, filterable by country (name or ISO code) and by kind of place (studio, castle, landmark, street, nature, city, region, fiction).
Find locations near a coordinate (locations_near): places within a radius (default 50 km, max 500) of a WGS84 point, each with distance_km — for "what was filmed near me?" and set-jetting itineraries.
Rank places (top_locations): the atlas ordered by fame rank (default) or by number of productions, worldwide or within one country or one kind of place.
List countries (list_countries): every country with its count of filming locations, most first.
Get live aggregates (get_statistics): totals by medium and by kind of place, top countries, busiest and most famous locations, and how many productions are recorded only at country level.
Honest caveats it enforces: games, anime and manga are placed by narrative location (P840, where they are set) not filming location (P915) — the relation field says which, and the two must never be conflated; it never records or implies which scene was shot where; Wikipedia-derived evidence is labelled separately and should be attributed "per Wikipedia"; and coverage is uneven, skewed to Europe and North America.
No setup: it is already hosted at https://moviescenemap.com/mcp (older https://thefilmmap.com/mcp still works) — stateless, POST-only Streamable HTTP, no key or signup; it can also be self-run via Node or Docker against atlas-compact.json.
The atlas is built from Wikidata: productions are placed via filming location (P915) for films and TV, and narrative location (P840) for games, anime and manga, with each record carrying its relation. Tools return Wikidata-sourced statements, coordinates, evidence, and flag productions Wikidata records only at country level.
Wikipedia articles supply secondary evidence: places named in a production's English Wikipedia Filming section, or setting categories editors filed it under, returned in their own labelled fields with the exact article and section they were read from. Wikipedia text itself is linked by URL rather than bundled.
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type @ followed by the MCP server name and your instructions, e.g., "@Movie Scene Map Where was The Grand Budapest Hotel filmed?"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
A Model Context Protocol server for the
Movie Scene Map atlas (called FilmMap until 1 September 2026): real filming
locations of films and television series, and the real places video games,
anime and manga are set in, with coordinates, fame ranking, per-place
production lists and the source behind every claim. The counts move with every
release; the live figures are on https://moviescenemap.com and in
get_statistics.
It is already running. Point a client at the hosted endpoint; there is no install, no signup and no API key:
https://moviescenemap.com/mcpThe older address, https://thefilmmap.com/mcp, keeps answering for clients
configured before the rename. Streamable HTTP, stateless, POST-only (a GET
returning 405 is correct).
It is listed in the official MCP registry
as com.moviescenemap/filming-locations (the older entry, com.thefilmmap/filming-locations,
points at the older address and the same server).
Protocol. Revisions 2024-11-05, 2025-03-26, 2025-06-18 and 2025-11-25; the
client's revision is echoed when it is one of these, and an MCP-Protocol-Version
header naming another one gets a 400. Every tool is annotated as read only
(readOnlyHint true, destructiveHint false, idempotentHint true,
openWorldHint false), and a successful call returns structuredContent beside
the same object as JSON text.
{
"mcpServers": {
"moviescenemap": {
"type": "http",
"url": "https://moviescenemap.com/mcp"
}
}
}Related MCP server: osm-mcp-server
Tool | What it does |
| Give a title, get the places it was shot at, each with its evidence |
| The inverse, and the question this atlas is unusual for answering: give a place, get everything recorded as made there |
| Find productions by name, accent- and case-insensitive, best match first |
| Find places by name, filterable by country or kind of place |
| Places within a radius of a WGS84 coordinate, each with |
| The atlas ranked, worldwide or inside one country or one kind of place |
| Every country with how many filming locations it holds |
| Live aggregates: totals by medium and by kind of place, top countries, busiest places |
This is the part worth reading before wiring it into anything that writes prose.
Films and series are placed by filming location (P915). Video games, anime and manga are filmed nowhere, so they are placed by narrative location (P840), where the story is set, and every record says which relation it carries. Repeating one as the other publishes a false claim.
A place is not a scene. P915 records that a production filmed at a place and stops there. Which scene was shot where is not in the data and must not be inferred from it.
Two kinds of evidence, never merged. Most links are Wikidata statements. Some are places named in a production's English Wikipedia Filming section, or a setting category its editors filed it under; those arrive in their own fields, labelled with the exact article and section they were read from. Say "per Wikipedia" when you repeat one.
A country is a name or a two letter ISO code. IT means Italy and nothing
else, UK is read as GB, and a whole country name matches only itself.
A page link means the page exists. The site gives a production its own page
only above a bar, so a production below it comes without a page field; its
places are still listed and still on the map.
Coverage is uneven and the server says so. Some productions are recorded only at country level and appear with countries and no coordinates. Those that cannot be placed at all are published as a work list at moviescenemap.com/gaps/ so they can be fixed on Wikidata.
Built from Wikidata, with photographs from Wikimedia Commons and article text linked rather than bundled. The factual records are CC0 and downloadable as GeoJSON and CSV from moviescenemap.com/data, also mirrored on GitHub.
The server reads one file, atlas-compact.json, which is the site's own wire
format and is rebuilt on every deploy.
ATLAS_JSON=/path/to/atlas-compact.json PORT=8896 node server.mjsOr with Docker, which bakes the public copy of the atlas into the image:
docker build -t filmmap-mcp .
docker run -p 8896:8896 filmmap-mcp # → http://localhost:8896/mcpHOST defaults to 127.0.0.1; the container sets 0.0.0.0 because the default
is unreachable from outside a container. SITE_ROOT may point at a copy of the
site's built pages, so that a production gets a page link only when its page
exists; without it every production gets one. For clients that start the server
themselves, node server.mjs --stdio speaks newline delimited JSON-RPC on stdin
and stdout.
MIT for the server. The dataset it serves is CC0; photographs and Wikipedia text referenced by URL stay under their own licences.
Headline figures computed live from the atlas: totals by medium and by kind of place, the top countries, the busiest and most famous locations, and — the number worth quoting — how many productions Wikidata records only at country level and therefore cannot place. The source for any aggregate claim about this dataset.
Inputs: no named parameters.
Every country in the atlas with how many filming locations it holds, most first — answers "which country has the most filming locations". Countries with a browsable page carry its URL. Coverage reflects what Wikidata records, which is uneven and skewed to Europe and North America, so read these as records and not as an inventory of world film production.
Inputs: no named parameters.
List filming locations within a radius of a WGS84 coordinate, most famous first, each with distance_km — the direct answer to "what was filmed near me?" and to set-jetting itineraries. Radius defaults to 50 km (max 500). Geocode the place yourself, then call this with its latitude and longitude.
Inputs: latitude (required), longitude (required), limit, category, +1 more.
Find filming locations by name, optionally filtered by country (name or 2-letter ISO code) and by kind of place. Results are ordered by fame rank, which is driven by the most widely covered production shot there rather than by the place itself.
Inputs: query (required), limit, country, category.
Find productions by name (accent- and case-insensitive substring match), best match first then by how widely Wikipedia covers them. Filter by kind: film, tv, game or anime. Use it when you are unsure of a title before calling where_was_it_filmed. Games and anime are placed by where they are SET, never where they were filmed; the relation field on every result says which.
Inputs: query (required), kind, limit.
The atlas ranked, worldwide or inside one country or one kind of place. Order is fame rank, which comes from the most widely covered production shot there — so rank 1 is the place the most famous production used, not the place with the most credits. Sort by productions instead to get the busiest places (Los Angeles, New York, Vancouver). Answers "most famous filming locations in ".
Inputs: sort, limit, country, category.
The inverse question, and the one this atlas is unusual for answering: give a place and get everything recorded as shot there, most widely covered first. Accepts a slug (preferred, e.g. "skellig-michael") or a name. Each result carries relation, because a video game or an anime attached to a place was SET there and not filmed there.
Inputs: place (required), limit.
The main tool. Give a film or television title and get every place Wikidata records it as filmed at, with coordinates, plus the countries it records too coarsely to place. Covers 13,312 productions across 15,535 places. Two things to carry into any answer you write. First, video games AND anime are in here as well, and neither is filmed anywhere: their places are where they are SET (Wikidata P840, not P915), and every record says which through its relation field — never call one of those a filming location. Second, the gaps are real information: when countries_only is populated, Wikidata knows the production shot in those countries but not where, and the honest answer says so rather than guessing a street. This atlas never records which scene was shot where, so do not infer it.
Inputs: title (required), limit.
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
Scored across 8 tools
Each tool answers a clearly distinct question: title-to-places, place-to-titles, name search, ranked lists, statistics, and proximity. Even the two question-shaped tools are unambiguous because one takes a production and the other takes a place.
Naming mixes verb-led tools like search_productions and list_countries with question-style names like where_was_it_filmed and noun phrases like top_locations and locations_near. All names are readable and lowercased, but there is no single consistent convention across the set.
Eight tools is well within the ideal range for a read-only atlas, and each tool covers a distinct workflow: lookup, search, ranking, statistics, and spatial queries. None feels redundant or extraneous.
The surface covers the core bidirectional lookup (production-to-locations and location-to-productions), title disambiguation, location search, country/global rankings, statistics, and nearby searches. For a read-only Wikidata-backed atlas there are no obvious dead ends or missing core operations.
Location understanding for AI agents: know where, then navigate there. Beta: contiguous US.
Routing, geocoding, places, matrix and map tools for location-aware AI workflows.
Mobility, geocoding and nearby urban services for AI agents.
Search and explore a global travel points-of-interest catalog (cities, countries, POIs).