Best IMDb MCP servers for agents in 2026

There is no official IMDb MCP from Amazon. For live title lookup, charts, and upcoming releases through a RapidAPI key, uzaysozen/imdb-mcp-server is the fullest local catalog: 23 tools, five rows a page, Python 3.13 or newer. For a hosted install with no key, Pipeworx serves four dump-backed tools at https://gateway.pipeworx.io/imdb/mcp, 50 anonymous calls a day or 200 with a free account. Personal ratings and a public watchlist belong to rollecode/imdb-mcp. Actor-credit analytics over a local 4.4 GB DuckDB copy of IMDb's non-commercial files belong to imdb-analytics-mcp. Trends MCP is first only for the live IMDb MOVIEmeter board: get_top_trends with type IMDb MOVIEmeter, default limit 100.

Which IMDb MCP server fits which job?

The right pick is the job, not a single winner. A title lookup server, a dump search, a personal ratings reader, and a live popularity board do not replace each other.

Trends MCP does not look up tt0111161, does not return cast lists, and does not read a personal watchlist. The IMDb MOVIEmeter API is a ranked board of titles. Pair a rank with get_growth on Google Search, YouTube, or Wikipedia. IMDb is not itself a get_growth source.

Is there an official IMDb MCP from Amazon?

Amazon has not published an IMDb Model Context Protocol server. The licensed path for commercial metadata is IMDb's AWS Data Exchange products, documented on developer.imdb.com. The free path for personal work is the non-commercial TSV dumps at datasets.imdbws.com, refreshed daily.

IMDb's help article "Can I use IMDb data in my software?" (help article G5JTRESSHJBBHTGX, accessed 12 Nov 2026) limits that dump path. Data must come from the published datasets, not from scraping imdb.com. Use is personal and non-commercial. Republishing a movie database is forbidden. Attribution required: Information courtesy of IMDb (https://www.imdb.com). Used with permission.

Every community MCP below is unofficial. Several are MIT-licensed as software while the IMDb files they read stay under IMDb's terms. Read those terms before hosting a shared endpoint.

What uzaysozen's RapidAPI server actually ships

uzaysozen/imdb-mcp-server is the local catalog most roundups mean by "an IMDb MCP." GitHub showed 22 stars and 8 forks on 12 Nov 2026, last push 28 Aug 2026, MIT license, Python 3.13 or newer. The README lists 23 tools against the RapidAPI IMDb listing at octopusteam-octopusteam-default/api/imdb236.

Search is one tool, search_imdb, with filters for title, genre, rating, year, and origin country. Title follow-ups are get_imdb_details, get_directors, get_cast, and get_writers. Chart tools include Top 250 movies, Top 250 TV, most popular movies, most popular TV, and a United States box-office list. Upcoming releases take a country code. Seven India-spotlight tools cover top-rated Indian, Malayalam, Tamil, and Telugu lists plus upcoming Indian and trending Tamil or Telugu titles. Helper tools return types, genres, countries, and languages.

Every collection endpoint pages at 5 items, with hasMore and nextStart. The in-memory cache is LRU, 100 keys, 10-minute TTL, swept every 5 minutes. HTTP timeout is 30 seconds. Default transport is stdio. TRANSPORT=http serves /mcp on port 8081.

Cursor and Claude Desktop install with uvx. Smithery ended free managed hosting in March 2026, so there is no shared remote instance in the README.

{
  "mcpServers": {
    "imdb_server": {
      "command": "uvx",
      "args": [
        "--from",
        "git+https://github.com/uzaysozen/imdb-mcp-server",
        "imdb-server"
      ],
      "env": {
        "RAPID_API_KEY_IMDB": "your_api_key_here"
      }
    }
  }
}

The README states a RapidAPI free tier is available. RapidAPI's HTML pricing table did not parse cleanly on this run, so check current RapidAPI pricing on the imdb236 listing before quoting a monthly dollar figure. HTTP 401 and 403 map to a missing or unsubscribed key.

This server wins live charts and cast lookup. It loses on personal ratings, on dump-scale SQL, and on MOVIEmeter ranks without a second connection.

When Pipeworx's hosted dump is the faster install

Pipeworx's IMDb pack is the no-clone option. The pack page (accessed 12 Nov 2026) describes title, ratings, crew, and episode data from the public non-commercial bulk dumps, ingested daily. No auth is required for the first calls.

Four IMDb tools ship on the pack:

The hosted URL https://gateway.pipeworx.io/imdb/mcp also lists shared gateway meta-tools (ask_pipeworx, discover_tools, and the rest). The README puts that extra set at roughly 30 schemas beside the four IMDb tools. @pipeworx/mcp-imdb 0.2.1 on npm (323 downloads in the week ending 12 Nov 2026, published 2 Nov 2026) is a local stdio bridge that lists only the pack's own tools unless PIPEWORX_META_TOOLS=1. GitHub pipeworx-io/mcp-imdb showed 1 star, last push 9 Nov 2026.

Pipeworx pricing (pipeworx.io/pricing, accessed 12 Nov 2026): 50 tool calls a day anonymous, 200 a day on a free GitHub-linked account. Past that, government and open data is $0.005 per call on pay as you go, $0.003 on the $49 month, $0.001 on the $99 month. IMDb sits on those dump files, so treat it as the open-data row, not a premium partner feed. Discovery calls (initialize, tools/list) do not draw the daily cap.

{
  "mcpServers": {
    "imdb": {
      "url": "https://gateway.pipeworx.io/imdb/mcp"
    }
  }
}

Pipeworx wins a five-minute hosted install. It loses on live page data that the dumps omit, on personal ratings, and on MOVIEmeter. The README itself marks the pack as unofficial and not affiliated with IMDb. Hosting dump-backed search for other people collides with IMDb's non-commercial rule; keep that in view.

Who should clone rollecode or the DuckDB analytics server

Two clone-only servers cover jobs the hosted packs skip. Both showed 0 GitHub stars on 12 Nov 2026. Star count is not the product. The data shape is.

rollecode/imdb-mcp v1.0.0 (released 25 Oct 2026, last push the same day) talks to the GraphQL API that imdb.com itself uses. No RapidAPI key. The unique tools are the personal ones: my_ratings, my_rating_for, find_my_ratings, my_rating_stats, my_watchlist, get_list, get_user_profile. Ratings history is fetched 250 rows at a time and kept for 15 minutes. Public catalog tools cover titles, people, credits, episodes, reviews, box office, Top 250, most popular, coming soon, and advanced title search. That is 21 tools on the README badge.

Setup is clone, uv venv && uv pip install -e ., then IMDB_PROFILE_ID=p.xxxx from a public ratings URL. Ratings and watchlist must be public. The README says the server is for reading a personal account and looking things up for that person. Do not put it in front of other people or a product, and do not store or republish what it returns. HTTP hosting, if used, sits on port 8620 behind nginx on 8621.

iamneilroberts/imdb-analytics-mcp is a different machine. uv run imdb-mcp update downloads the official dumps and builds data/imdb.duckdb at about 4.4 GB. Six tools: rank_people, search_people, get_person, compare_people, find_collaborations, search_titles. Status as of 9 Sep 2026: ingest, analytics, and MCP tools are built. Wikidata enrichment for nationality, gender, and awards is M4; those filters still error.

The 9 Sep 2026 definition-of-done chain is the honest sample. rank_people for the 25 people with the most acting credits returned Brahmanandam at 1,131 credits, then Eric Roberts at 725, against 2,380,714 people with at least one credit. find_collaborations in degrees_of_separation mode caps at three hops. IMDb's principal-cast list holds at most 69 people per title, so a missing shared credit is not proof two people never worked together. search_titles notes that only 1.7 of 12.8 million titles carry any rating, so a rating filter is also a coverage filter.

MIT covers the code. IMDb forbids republishing the files as a movie database. Run it locally. Do not host it as a shared service.

tmdb-mcp 0.14.0 (npm, 218 weekly downloads, published 22 Oct 2026; GitHub Grinv/tmdb-mcp, 5 stars, last push 22 Oct 2026) is a TMDB server that can attach IMDb, Rotten Tomatoes, and Metacritic scores through OMDb. Runtime is Node 22.18 or newer. CACHE_TTL_MS defaults to 300000. TMDB_MIN_INTERVAL_MS defaults to 60. HTTP_TIMEOUT_MS defaults to 15000, with 2 retries. find_by_imdb_id resolves a tt id to a TMDB record. OMDb is optional: the free key form on omdbapi.com (accessed 12 Nov 2026) prints a 1,000 daily limit. Without that key, TMDB metadata still answers and the ratings field reports unconfigured. This is not an IMDb catalog. Use it when TMDB search, watch providers, and trailers are the job and an IMDb score is enrichment.

How Trends MCP ranks films on MOVIEmeter

Trends MCP wins the live popularity board, not the title page. The MCP schema on 12 Nov 2026 names the feed IMDb MOVIEmeter. get_top_trends takes no keyword. Default limit is 100 (10 MCP requests at 1 request per 10 rows). Max limit is 200. REST still bills one successful call as one request. Sort defaults to current rank. sort=rank_change with window 1d, 3d, 7d, 14d, or 30d returns climbers against a prior snapshot.

A pagination pull stored on 25 Aug 2026 returned count 100 on IMDb MOVIEmeter even when limit was 200. Offset 100 was empty. Treat 100 as the board depth from that pull, not a promise that every future day fills 100 rows. Read count and as_of_ts. The board is ranks and titles. It does not expose star ratings, vote counts, cast, or box office.

The Cursor object needs transport http. Homepage copy-config on 12 Nov 2026 still omitted that field. The working block, matching Trends MCP for Cursor, is:

{
  "mcpServers": {
    "trends-mcp": {
      "url": "https://api.trendsmcp.ai/mcp",
      "transport": "http",
      "headers": {
        "Authorization": "Bearer YOUR_API_KEY"
      }
    }
  }
}

Ask for the board with a low limit first. A default 100-row page costs 10 MCP requests. Ten rows cost 1. Type strings and billing live on the get_top_trends page; some older pages still print a default limit of 25, which does not match the live MCP schema fetched today.

llms.txt (Updated 12 Nov 2026) lists 54 data sources across 27 platforms and five billed tools: get_time_series, get_growth, get_top_trends, get_context, get_reviews, plus free get_usage. Plans: Free $0 / 100 requests a month; Starter $19 / 1,000; Pro $49 / 5,000; Business $199 / 25,000. Yearly cards: Starter $16 / $192, Pro $39 / $468, Business $159 / $1,910. Failed calls are free. MCP get_top_trends is 1 credit per 10 rows; REST is 1 call = 1 request.

A title that climbs MOVIEmeter still needs an off-IMDb check. Shorten The Odyssey (2026) to The Odyssey before get_growth. Parenthetical years break many Wikipedia and Google lookups. For search-interest context beside the board, see IMDb MOVIEmeter vs Google Trends.

What IMDb's dataset terms allow

The dump-backed servers (Pipeworx, imdb-analytics-mcp) sit on files IMDb publishes for personal, non-commercial use. The GraphQL server (rollecode) sits on the same site-facing API the website uses, with an explicit "do not productize" warning in its README. The RapidAPI server (uzaysozen) sits on a third-party listing, not on Amazon's licensed IMDb API.

That split decides hosting. A local stdio process for one analyst is the shape IMDb describes for dumps. A public MCP URL that answers other people's title searches is a different fact pattern. Trends MCP's MOVIEmeter board is a ranked popularity feed through Trends MCP's own API, not a republished title database.

None of this is legal advice. Read G5JTRESSHJBBHTGX and the license block inside each TSV before shipping.

FAQ

Is there a best overall IMDb MCP server?

No. uzaysozen wins live lookup and charts if a RapidAPI key is acceptable. Pipeworx wins a hosted dump search with no key. rollecode wins personal ratings. imdb-analytics-mcp wins actor-credit SQL on a local DuckDB file. Trends MCP wins MOVIEmeter ranks and the cross-source growth follow-up. Connecting two of them is normal.

Does Trends MCP return IMDb ratings or cast?

No. get_top_trends type IMDb MOVIEmeter returns ranks and titles. Star ratings, vote counts, cast, and box office are not on that board. Use uzaysozen, Pipeworx, or tmdb-mcp plus OMDb for those fields, then optionally grow the shortened title on Google Search, YouTube, or Wikipedia.

Can an IMDb MCP run in Cursor without a local install?

Yes for Pipeworx (url https://gateway.pipeworx.io/imdb/mcp) and Trends MCP (url https://api.trendsmcp.ai/mcp with transport http and a Bearer key). uzaysozen, rollecode, and imdb-analytics-mcp start as local processes. tmdb-mcp can run through npx -y tmdb-mcp with a TMDB v4 Read Access Token.

What does IMDb allow for non-commercial dump use?

Help article G5JTRESSHJBBHTGX, accessed 12 Nov 2026, allows limited personal, non-commercial use of the published datasets, forbids scraping imdb.com for that purpose, and forbids republishing a movie database. Attribution: Information courtesy of IMDb (https://www.imdb.com). Used with permission. Amazon can withdraw permission. Check current terms before hosting.

How much does a MOVIEmeter pull cost on Trends MCP?

On MCP, 1 request per 10 requested rows, so limit 10 is 1 request and the default limit 100 is 10. On REST, one successful call is one request even at limit 200. Free includes 100 requests a month. Starter is $19 for 1,000. Check current pricing on the site; failed 4xx calls do not count.

Sources (all accessed 12 Nov 2026): GitHub uzaysozen/imdb-mcp-server README; Pipeworx IMDb pack and pricing pages; GitHub pipeworx-io/mcp-imdb and npm @pipeworx/mcp-imdb 0.2.1; GitHub rollecode/imdb-mcp; GitHub iamneilroberts/imdb-analytics-mcp; npm tmdb-mcp 0.14.0; omdbapi.com key form; IMDb help article G5JTRESSHJBBHTGX; Trends MCP llms.txt (Updated 12 Nov 2026) and live MCP get_top_trends schema.