Methodology

Agent Ready tracks every plugin and connector across the biggest AI app stores: ChatGPT, Claude, Cursor, Grok Bot and Meta's Muse. We scan and update our data daily at 06:30 ET.

What we track

For each plugin or connector, we record all the available data: name, description, publisher, tools, sign-in method and more. We also track changes over time. You can subscribe to get daily updates:

None of the stores publish usage data: no install counts and no active user numbers. However, ChatGPT publishes a directory rank, and Claude publishes a 'popularity rank' and a 'trending rank.' We track these to understand changes over time.

Quick vocab: defining key terms

A 'connector' lets your AI assistant use data and tools from a third party app, via an open standard called the Model Context Protocol. An 'MCP server,' a 'ChatGPT app' and a 'connector' are the same thing; we'll just use 'connector' where possible.

A 'skill' is a set of written instructions the assistant loads for a task. It's like a recipe, or a standard operating procedure.

A 'plugin' is a bundle of 'skills' and 'connectors' that you can add to your AI tool all at once.

123

Skill

A recipe

Instructions that teach AI to do a task your way.

+

AI

Connector

Also called an MCP server

Lets AI read and write data in your other tools.

=

+

Plugin

Skills plus connectors

One install gives AI the steps, the data and the tools.

Not every plugin has a connector, and not every connector is part of a plugin, and the whole thing is terribly confusing.

How we group listings into apps

Spotify has a listing in ChatGPT, one in Claude and one in Muse. We show it as one app with three store ranks. We match listings by owner (the organization or website domain) and by name, so "Notion" and "Notion MCP" from the same company end up together.

Sometimes community members make unofficial connectors for their favorite tools, so we're careful to try to keep those separate from official ones. A listing that lives on GitHub joins a brand's app only if the store or the brand published it, so a community plugin that borrows a brand's name is listed separately wherever possible. Sometimes we end up with duplicates, but we would rather play it safe.

What the ranks mean

Only ChatGPT and Claude publish data on connector/plugin rank, and they measure different things.

ChatGPT's rank is the position in its plugin directory. OpenAI has said that this order, and which plugins ChatGPT suggests in a chat, is deterministic and based on conversion, active usage and retention.

Claude's rank is Anthropic's popularity rank. Its connector directory docs say plainly that usage rank is based on the "number of distinct Claude accounts that sent your server any MCP message in the last 30 days." Claude also publishes a separate trending rank for connectors that are growing fast, and we mark its trending top 100 with ▲.

We never mix them: #20 in ChatGPT isn't better or worse than #30 in Claude. Cursor, Grok Bot and Muse don't share any public rankings, so those apps just show as 'listed'.

What the tools bar shows

Every tool an app offers either reads data, writes it, or deletes or changes it in a way you can't undo. The bar shows the split of tools by type. ChatGPT requires developers to label every tool, so those shares come from the store. For other stores we estimate from tool names.

More tools isn't better. A connector with two well-described tools can outrank one with two hundred, and the bar shows proportions, not size.

Official and community

We call an app official when the store vouches for the publisher (ChatGPT's reviewed developers, Claude's Anthropic and partner tiers, Cursor's verified publishers, every Muse listing), when the store built it, or when its MCP server runs on the brand's own domain. Everything else the stores list without verification is 'community'.

How we assign categories

The stores label the same app differently, so we use one shared taxonomy across all four. Each app gets one category for its main job, an audience (consumers, business or developers) and, for work tools, an industry. The labels come from reading each app's tools and description, and new listings are tagged every day.

For agents

Every page is server-rendered HTML. There's a plain summary at /llms.txt, every app is in the sitemap, and the Work with us form is also a WebMCP tool that browser agents can call.