Introducing threefold: market-entry answers for app builders and their agents
Taking an app into a new language market is mostly a series of small, specific questions. What do the top apps in this niche look like there? Did the ones that win at home cross over, or did someone local take the spot? Is their listing translated, or written for the market? Which Spanish are they using? And, closer to home: is my own app actually ready, or does it just have an es.json file?
threefold answers those questions for app builders and for the agents they work with. Everything in it is callable as a tool from Claude, ChatGPT, Grok, Claude Code, Codex, Cursor, VS Code or Gemini CLI, and the same endpoints are available as a REST API, a CLI and typed TypeScript and Python clients.
#Two kinds of question
threefold covers two families of question.
What is different there. The core is a cross-market comparison: how apps like yours show up, get talked about and advertise in the target market compared with your own. We split that into three pillars. Brand is how an app presents itself in each storefront: title, copy, screenshots, declared languages, rank, reviews. Paid is what it advertises, from ad libraries. Organic is what creators post about it.
Is my app ready for there. The i18n readiness check runs on your machine against an iOS, Android, web, Flutter or React Native project and returns a verdict per target market. Market resources then help fix what it finds: the terms top listings in that market actually use, variant lint for Spanish, a store listing check, price points and a launch checklist.
#What is in the library today
The library started from an App Store snapshot generated on 27 October 2026: 10 markets (the UK, the US, Spain, Mexico, Argentina, Brazil, Germany, France, Japan and South Korea), 6 niches, the top 10 free apps per niche in each market, and every charting app's listing in every one of those markets. That is 360 apps and 3,600 storefront listings. Live ingest adds Google Play listings, store reviews and history: every listing change, every day's chart position.
Each market carries a profile, because the same language is not the same market. Mexico and Argentina are tagged es-MX and es-AR, not just es. Japan's market-native channels are LINE, X and TikTok. Japan's monetisation note reads "Highest-value payers. US subscription prices are a poor default."
Brand data needs no supplier keys. Paid and Organic do, and those supplier connections are not live yet. Until they are, answers about ads and creator posts are live links to each platform, filtered by country, and the endpoint descriptions say so. We would rather an agent tell you "this is a link, not data" than present an empty table as a finding.
#Built for agents
Most of the design follows from one assumption: the first reader of a threefold answer is often a model.
Over MCP, threefold provides four unified tools (app, market, check, workspace; see Agent tools). Every callable REST endpoint is described in the registry with its task in plain language, its inputs and examples, the markets and stores it covers, where its data comes from, its known limits and its cost class, searchable via GET /api/discover.
Results are typed. MCP tools return a concise markdown digest under a strict token budget with direct citation links for the agent and a threefold.eco link for the user, alongside typed structured content. Every response carries its cost, on HTTP as X-Threefold-Cost-Micro-Usd and on MCP in _meta. Errors are a closed set, and each says whose fault it was and whether retrying will help. When a call runs into a plan limit or a spending cap, the answer is a 402 with a specific code, and the agent skill we ship tells the agent to stop and ask you rather than look for a way around it.
You can try it without an account. Library reads work keyless up to a daily cap per IP address, over REST or over MCP with ?keyless=1. Sandbox keys (tf_test_…) return fixture data, cost nothing and never call a supplier, so you can put threefold in a test suite.
#Visuals and watches
Store listings are mostly pictures, so threefold returns them as pictures. Tools return small previews the model can look at, plus links to cite and download, and a threefold.eco link so you can inspect the visual evidence in your browser.
Watches follow an app or a niche in the markets you choose and send a signed webhook or an email when a listing, its screenshots or its price change, when it moves in the charts, or, once Paid data is live, when it starts a new ad.
#Where it runs
The API and worker run in the Netherlands and the database in Frankfurt, and the server refuses to start if its database or storage is configured outside the EU. The i18n check never uploads your code; if you choose to share a report, only the report is sent.
#What is not finished
We would rather list it than have you find it:
- Paid and Organic collection waits on supplier accounts.
- Prices are in beta. The plan structure and limits are live; paid plan prices are not published, and the per-call price is a placeholder until it is calibrated against real agent sessions.
- Chat-app sign-in passes our own in-process tests of the full OAuth flow, but we have not yet recorded a test inside each chat app, and threefold is not yet listed in their directories.
- The
threefoldnpm package and the TypeScript and Python clients are not published yet. The connect page has the exact settings to use in the meantime.
#Start
Add https://api.threefold.eco/mcp to your agent and ask it about a market you are considering. The connect page has settings for each client, and the docs cover the rest. If threefold does not cover a market, app or channel you need, have your agent call request_coverage: those requests decide what we add next.