How the rankings are determined: combining downloads, stars, and on-site popularity
When choosing an MCP server, Skill, or agent, rankings can help us quickly discover projects that have attracted attention, but “popularity” does not mean “right for you.” Downloads usually reflect the scale of actual trials or use, while stars show user approval and willingness to bookmark; on-site popularity can serve as a reference for interest within the directory. These signals represent different things and should not simply be added together and treated as a ranking of project quality.
Ideally, compare the time range and measurement criteria for each metric first, then assess them in light of the project’s intended use. For example, even if a Skill for a specific scenario serves a narrower audience, it may still be a great fit for the relevant workflow; and even a highly popular MCP server should still be checked for the data it provides, its operation permissions, and usage limitations.
The on-site materials available for this article include 6 entries and provide their on-site popularity, but do not include download counts, star counts, or complete data for a top-ten list. So the article below introduces these 6 real entries and treats popularity as a limited reference, rather than calling them the “top ten based on combined downloads and stars.” The three most popular Skills are upstash-redis-js, okx-dex-market, and story-long-scan.
Popular projects at a glance: Redis, DEX market data, and web novel ranking Skills
In terms of use cases, these three Skills serve application development, crypto market research, and online fiction content analysis, showing that “popular” does not correspond to just one type of user.
| Entry | Type | On-site popularity | Primary use |
|---|---|---|---|
| upstash-redis-js | Skill | 20668 | Use the TypeScript／JavaScript SDK to work with Redis |
| okx-dex-market | Skill | 18853 | Query DEX tokens, markets, and related signals |
| story-long-scan | Skill | 16289 | Analyze online fiction platform rankings and new-book data |
For developers, upstash-redis-js describes the HTTP-based @upstash/redis client, suitable for scenarios including Next.js, Vercel, Cloudflare Workers, edge runtimes, and Node.js. The listed uses include caching, session storage, key-value storage, counters, distributed locks, and implementing leaderboards and rankings with sorted sets. When evaluating it, first compare your runtime environment and data access needs, then confirm whether it fits your existing architecture.
okx-dex-market is for DEX market research. It covers information such as token searches and rankings, liquidity, holders, prices, and candlesticks, as well as smart money, KOL, whale signals, and token sentiment. It suits research workflows that need to query multiple types of market information in one place; however, “queryable” does not mean investment advice, and users should independently verify the data and assess the risks.
story-long-scan focuses on online fiction platform rankings and new-book data, helping identify market trends and popular genres. If your work involves tracking topics, content research, or following new releases, this kind of tool may be more direct than a general-purpose data MCP. It offers trigger phrases such as /story-long-scan and /长篇扫榜, which you can try in your actual workflow.
On-chain data MCPs: three.ws Activity, Coin Clash, and Intel
All three on-chain entries come from three.ws, but each serves a different purpose. In the provided materials, all three have an on-site popularity score of 241; the same popularity score only means that this metric is the same in the recorded materials. It does not imply that their downloads, star counts, or usage scales are also the same.
- three.ws Activity: View trending agents and tokens, the $THREE holders leaderboard, and site-wide activity.
- three.ws Coin Clash: View the Coin Clash faction battle dashboard and leaderboard, join a faction, or rally supporters.
- three.ws Intel: Track smart money scores for tokens, wallet credibility, signal feeds, and KOL rankings.
These three tools provide different entry points to information: Activity focuses on browsing updates and rankings, Coin Clash offers a gamified experience with faction interactions, and Intel focuses on signals and wallet-related information. When choosing, first clarify whether you want to “view updates,” “take part in interactions,” or “research signals,” then check whether the actual data returned meets your needs. In particular, wallet credibility and smart money scores should be treated as analytical clues, not guaranteed conclusions.
Choosing by use case: development, market research, and community interaction
If your main task is application development and infrastructure integration, start with upstash-redis-js: it covers common needs such as caching, sessions, key-value storage, and leaderboards. Check that the SDK matches your runtime environment and that read/write patterns, expiration policies, and other details align with your application design.
If your goal is market research, choose based on the question: okx-dex-market provides broad DEX market and signal queries; three.ws Intel focuses on smart money, wallet credibility, signal feeds, and KOL rankings; three.ws Activity is better suited to quickly browsing activity and holder rankings. The entries cover different subjects, so it’s best to validate the results with a specific question rather than deciding solely by ranking position.
If your work involves online fiction content and genre research, story-long-scan is a closer fit; if you want to take part in community-style interactions, explore three.ws Coin Clash. For these two types of use, the user experience and fit with your workflow are often more important than a single popularity score.
What to keep in mind when reading rankings: differences in metrics and update frequency
When reading rankings, use the following questions as a checklist:
- What does each metric represent? Downloads, stars, and on-site popularity are defined differently and should not be treated as interchangeable without verification.
- Are the reporting periods consistent? Data from different time windows cannot be compared directly, and rankings may change as they are updated.
- Are the entry types the same? MCP servers and Skills may differ in their purpose, invocation methods, and user base, so rankings across types are not like-for-like comparisons.
- Does popularity align with your task? High visibility can help you discover candidates, but it cannot replace checking their functionality, permissions, and data quality.
- Is there enough data to support a conclusion? The information available in this article covers the on-site popularity of 6 entries, not complete top-ten data including downloads and stars. Therefore, it would be unwise to claim that any entry has an overall ranking based on this information.
When choosing a tool, you can first use the ranking to narrow down your options, then try them in your own scenario and verify their functionality. To compare the “top ten,” use download and star data based on the same reporting period and clearly defined metrics. Until those data are available, treat rankings primarily as a discovery starting point, not as a definitive conclusion.