Points from each chart
Each day an app earns points on every chart its store orders by popularity: 1 divided by its position. The top of a chart counts far more than the middle, and nothing below the chart's cap counts at all. An app's points in a store are the mean of those daily points over the last 28 days.
- Σ
- Every counted chart, on every day of the last 28.
- w
- The chart's weight: 1 for a main popularity chart, 0.5 for a curated list, a home shelf or a subchart.
- p
- The app's position on that chart that day. A position past the chart's cap (50 or 100) earns nothing.
- D
- The days in the window the store has chart data for. A day the app is on no counted chart adds 0.
| Position | Points |
|---|---|
| #1 | 1 |
| #2 | 0.5 |
| #3 | 0.333 |
| #5 | 0.2 |
| #10 | 0.1 |
| #25 | 0.04 |
| #50 | 0.02 |
Each store's apps are ranked by those points. Points from different stores don't compare: ChatGPT counts 13 category lists, 3 curated lists and its home shelves, and Grok Bot one list, so 0.4 means something different in each. Ranks do compare, and the score is built from them.
Which charts count
Only charts the store orders by popularity count. ChatGPT's category lists follow popularity through about #50 and launch order below it, so they stop counting at #50. Launch-ordered lists, such as Claude's New, count nowhere.
| Store | Counted charts | Weight | Cap |
|---|---|---|---|
| ChatGPT | The 13 category lists, such as Productivity and Finance | 1 | 50 |
| ChatGPT | Popular, New & Noteworthy, Small Business and the home shelves | 0.5 | 50 |
| Claude | Most popular and Connectors by popularity | 1 | 100 |
| Claude | Most popular's tag lists | 0.5 | 50 |
| Cursor | The directory's Top list | 1 | 100 |
| Cursor | Featured plugins | 0.5 | 50 |
| Grok Bot | Popular | 1 | 100 |
The score
An app gets one score. Each store that ranks it adds 21 / (20 + its rank there), through its best listing. #1 is worth 1, and #1 in one store is worth about the same as #20 in two, so one big win doesn't outrank several good ranks. Grok Bot lists Cursor's marketplace, so it counts as Cursor: a plugin on both scores once, at the better of its two ranks. A store isn't weighted by its size, because no store publishes its audience to weight it by.
- Σ
- Every store that ranks the app.
- r
- The app's rank in that store, from its best listing there.
- L
- How many of Perplexity and Muse show the app's own connector.
Perplexity and Muse count by listing, not by order. Perplexity's order is its own catalog's and Muse's is alphabetical, and neither helped the checks below. Each adds 0.1 when it shows the app's own connector. A connector that Pipedream or Merge runs for Perplexity adds nothing, because a third party's connector says nothing about the app.
| Store | Rank | Adds |
|---|---|---|
| Claude | #1 | 1 |
| Cursor | #2 | 0.955 |
| ChatGPT | #3 | 0.913 |
| Perplexity | Listed | 0.1 |
| Muse | Listed | 0.1 |
| Score | 3.07 |
The board ranks apps, connectors, extensions and MCP servers. It leaves out Claude Code plugins and Cursor rules, and numbers the listings left in each store 1, 2, 3 among themselves.
Tier and momentum
- Tier is where an app stands today: Top 10 on any counted chart, Top 50, Charted, meaning on a counted chart at some point in the 28 days, or Listed, meaning only on Perplexity or Muse. Across stores it is the best tier of the app's listings.
- Momentum is the score minus the score 28 days earlier. It shows once there is an earlier window to compare with, and only ranks move it.
- Days charted and peak, on a store's board, are the days in the window with a counted position and the best position in it.
Checked against installs
A chart position isn't a count of users, so we test the rank against the counts some stores publish: Claude's installs and popularity, and cursor.directory's installs. For each count, the score is computed from the other stores only and compared with the count by Spearman correlation. Grok Bot lists Cursor's marketplace, so it is left out with Cursor.
| Held-out count | Score from the other stores | Best single other store |
|---|---|---|
| Claude installs | 0.59 | 0.57 |
| Claude popularity | 0.59 | 0.46 |
| cursor.directory installs | 0.30 | 0.31 |
It beat the best single other store on Claude installs and Claude popularity and came within 0.01 of it on cursor.directory installs. The site has ranked apps by this score since November 2026.
scores_methods gives the definition and validation, and scores_apps the board across stores (API docs). A change to the definition gets a new version, computed beside the old one. The site shows oa_rank@3.