Best App Store MCP servers for agents in 2026

Apple does not publish an App Store MCP. Third-party servers split by job. For live iOS Top Free and Top Paid charts plus App Store reviews in Cursor or Claude, Trends MCP is the first connect: get_top_trends types App Store Top Free and App Store Top Paid default to 100 rows, and get_reviews source App Store matches an app name or numeric id. Sonar wins keyword difficulty and both-store ASO (80 tools, 30 free requests per IP per day). Appfigures wins linked-app downloads and revenue. AppFollow wins review replies on tracked apps. Local packages @perttu/app-store-mcp and app-market-mcp win keyless lookups on Apple's public endpoints, with Apple's search cap near 20 requests per minute. None of them submit builds to App Store Connect.

Which App Store MCP should an agent connect first?

Connect the server that answers the question the agent is about to ask. Live chart rank is a different object from keyword difficulty, and both are different from last month's proceeds.

Trends MCP is first only for live multi-source trend data that includes those two App Store boards. It is not the best overall MCP, and it is not an ASO suite. Sonar, Appfigures, and AppFollow beat it on keyword volume, own-app actuals, and reply workflows.

Does Apple ship an official App Store MCP?

No. Apple's public surface is the App Store Connect REST API for the apps a developer already owns. The overview at developer.apple.com/app-store-connect/api (accessed 14 Nov 2026) describes workflow automation for metadata, TestFlight, and signing. It does not list an MCP server.

That split explains the two product families people mix up in search. Public-store MCPs read listings, reviews, and charts for any app. Connect MCPs hold a .p8 key and can submit versions. This roundup is the first family. Connect servers such as apple-app-store-connect-mcp belong in a publishing setup, not in a trend-research client.

How does Trends MCP expose App Store charts and reviews?

Two tools, two contracts. Charts are get_top_trends with no keyword. Reviews are get_reviews with a name or numeric App Store id. Android history is a third tool, and it refuses iOS ids.

The live catalog on 14 Nov 2026 lists App Store Top Free and App Store Top Paid as daily boards. country on get_top_trends applies to Google Trends, Pinterest, TikTok hashtags, X, and YouTube. It does not apply to those App Store boards. Review pulls are the ones that take a store country, defaulting to the United States.

Omit limit on get_top_trends and the server bills for 100 rows (10 requests, because MCP charts cost 1 request per 10 rows). Drop limit to 10 when the job is a snapshot, not a dump.

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

Cursor wants url plus transport http on https://api.trendsmcp.ai/mcp. Claude and ChatGPT sign in at the www host /mcp path instead. The same key also POSTs to https://api.trendsmcp.ai/api. Full Cursor JSON lives on the Trends MCP server for Cursor page.

A chart call that costs one request:

{
  "mode": "get_top_trends",
  "type": "App Store Top Free",
  "limit": 10
}

get_reviews source App Store matches 'notion' or '1232780281'. Default limit is 10 reviews, default days is 90, max limit 20, max days 365. source all is three review stores (App Store, Google Play, Amazon) and costs 3 requests.

Android history is the trap. get_growth sources App Downloads and App Rankings take an Android bundle ID only (com.openai.chatgpt). A display name, an iOS numeric id, or an iOS bundle that is not on Play returns a miss. The App Store Top Free API page walks that split. Pairing both stores in one agent is the App Store intelligence in Cursor workflow.

{
  "mode": "get_growth",
  "source": "App Downloads",
  "keyword": "com.openai.chatgpt",
  "percent_growth": ["3M", "12M"]
}

llms.txt dated 14 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. App Store is one of those platforms, not a standalone product. The same key can follow a chart title onto Google Search, TikTok, or Amazon without a second MCP.

What local App Store MCP servers run without a key?

Three MIT packages scrape or call Apple's public endpoints from stdio. They win when the agent should not hold a SaaS login. They lose when Apple's search endpoint throttles, and they lose hosted clients that cannot spawn npx.

@perttu/app-store-mcp 0.1.0 (npm, created 12 Aug 2026, 1 GitHub star) exposes eleven read-only tools. app_store_reviews pages 1 through 10. app_store_search accepts num 1-200. app_store_list can set fullDetail: true, which fires one extra request per chart row; the README tells agents to keep num small. Default country is the United States storefront. Node 20 or newer.

{
  "mcpServers": {
    "app-store": {
      "command": "npx",
      "args": ["-y", "@perttu/app-store-mcp"]
    }
  }
}

app-market-mcp 0.1.0 (npm, 7 GitHub stars, 1 fork) is the market-level counterpart. share_of_search returns a top-50 leaderboard plus studio concentration. keyword_matrix is capped at 10 keywords × 10 apps. market_map returns free/paid mix and median age in one call. Watchlists store JSON under ~/.app-market-mcp/. The README states Apple's search endpoint soft-limits around 20 requests per minute per IP, and the server paces search calls about 3 seconds apart, so a 10-keyword matrix takes about 30 seconds. "Unranked" means not in the top 200.

appstore-mcp-server on PyPI is 0.0.8 and requires Python 3.12+. The README (fetched 14 Nov 2026) requires macOS 26 or newer on Apple Silicon. Eleven MCP tools include search_ranked against the MZStore API so results match App Store ranking, plus top_charts, find_app_rank, and analyze_keyword with a 0-100 competitiveness score. CLI top defaults to 25 rows (range 1-200). The GitHub repo description still said "MCP Server (coming soon)" on 14 Nov 2026, which contradicts the published tools. Skip it on Linux agents and on Intel Macs.

Apple can change those public endpoints without notice. @perttu/app-store-mcp does not cache or throttle; the client owns request rate. app-market-mcp throttles on purpose. Neither returns keyword search volume. That number is a paid ASO object.

Which hosted ASO servers beat Trends MCP on keywords?

Sonar, Appfigures, AppFollow, and Manifold. Each one wins a slice Trends MCP does not ship.

Sonar's docs (accessed 14 Nov 2026) count 80 tools covering the REST API, 79 on the hosted endpoint because screenshot export to disk is local-only. npm latest for @sonarapp/mcp is 0.13.2 (modified 14 Oct 2026). The same docs "at a glance" block still printed @sonarapp/[email protected]. Free mode needs no key: seven stateless endpoints share 30 requests per day per IP, and sonar_keyword_metrics allows 5 keywords per day. Signup adds 50 credits. Prepaid packs start at $10 for 1,000 credits and never expire. Indie is $29/month (10 apps, 2,500 tracked keywords, 1,000 API requests/day). Agency is $149/month. Hosted URL: https://trysonar.app/mcp. Keys start with aso_. Keyword research and rank tracking cover 72 markets; lookup, search, reviews, and ASO scores cover 175 App Store storefronts and 145 Play markets.

Appfigures MCP is https://mcp.appfigures.com/mcp. Linked-app downloads, revenue, subscriptions, ratings, reviews, and usage have no extra data charge. Public competitor estimates need the Boost plan at $599.99/month plus the Public Data API add-on. Estimate calls cost 30 credits; category ranks and keywords cost 10; product metadata costs 2. Every account includes 1,000 API requests per day (help article updated 27 Oct 2026). Connect is $9.99/month on the public pricing page. Starter tracks up to 5 apps at $0.

AppFollow MCP is https://mcp.appfollow.io/mcp (support article 49071302595601, updated 12 Nov 2026). Any role can connect. Owners and Admins get write access, including review replies. Other roles are read-only. One connection covers up to 9 workspaces. The session lasts 30 days. MCP is not a separate SKU; calls spend the workspace's API credits. Configure calls are free, Get calls are charged per read, Do calls are charged per item or batch. Quote dollars from AppFollow's own pricing page before locking a budget. The MCP layer itself adds no extra fee.

Manifold's App Store page (accessed 14 Nov 2026) lists apps_get_app and apps_get_reviews from 1 credit, apps_get_keywords and competitor tools from 5, apps_search_apps from 31, and apps_get_top_charts from 1. Reviews cost 1 credit per 100, up to 500 a call. Listings, reviews, and charts follow the storefront the agent passes. Keyword and competitor data covers United States English only. Free is 250 credits a month, no card. Pro is $29 for 1,500 credits or $49 for 3,000. Cache hits cost 25% of the live price. Manifold does not return own-app downloads; that remains App Store Connect or Appfigures.

Sensor Tower, AppTweak, and MobileAction appear in other App Store MCP roundups as API-backed market-intel servers. Those are enterprise ASO APIs with MCP wrappers, not the first connect for a live Top Free board. Check current pricing on each vendor before treating a quote as current.

FAQ

Is there an official Apple App Store MCP server?

No. Apple publishes the App Store Connect REST API for apps a developer already owns, not an MCP server that reads the public store. Public chart, review, and keyword servers are all third-party. Connect MCPs that hold a .p8 key are a publishing tool, not a trend feed.

Which App Store MCP is best for Cursor?

For live Top Free / Top Paid charts and App Store reviews next to the repo, Trends MCP. Paste the trends-mcp object with transport http and a Bearer key, then fully quit Cursor. For keyword difficulty, add Sonar. For last week's proceeds on linked apps, add Appfigures. One server rarely covers all three jobs.

Can Trends MCP return iOS download counts?

No. get_growth sources App Downloads and App Rankings take an Android bundle ID. An iOS numeric App Store id or a display name is the wrong keyword. iOS chart rank is get_top_trends type App Store Top Free or App Store Top Paid. iOS review text is get_reviews source App Store.

How much does an App Store MCP cost to try?

Trends MCP Free is 100 requests a month with no card. Sonar free mode is 30 requests per IP per day. Manifold Free is 250 credits a month. @perttu/app-store-mcp and app-market-mcp are $0 beyond Apple's public-endpoint throttling. Appfigures Starter tracks 5 apps at $0. AppFollow MCP runs on the Free plan with that plan's API credits; confirm the current allowance on AppFollow's pricing page.

Do local App Store scrapers replace a hosted ASO MCP?

No. They return public listings, reviews, and charts. They do not return keyword search volume, difficulty, or estimated revenue. @perttu/app-store-mcp search maxes at 200 results. app-market-mcp treats anything outside the top 200 as unranked. Hosted ASO servers exist because those volume fields are not on Apple's public search endpoint.

Sources: Apple App Store Connect API overview (accessed 14 Nov 2026); Trends MCP llms.txt and data-catalog.md (Updated 14 Nov 2026); Sonar MCP docs; npm @sonarapp/mcp 0.13.2; Manifold App Store MCP and Manifold pricing; Appfigures MCP, Appfigures pricing, and Appfigures help article 1seiibo (updated 27 Oct 2026); AppFollow support article 49071302595601 (updated 12 Nov 2026, accessed 14 Nov 2026); GitHub READMEs for plahteenlahti/app-store-mcp, EYamanS/app-market-mcp, and drewster99/appstore-mcp-server; npm @perttu/app-store-mcp 0.1.0, app-market-mcp 0.1.0; PyPI appstore-mcp-server 0.0.8; GitHub API star counts on 14 Nov 2026.