carrot is one README file listing free AI sites, and its markers are not filterable columns
A directory of AI tools to help you quickly find free, practical, and efficient website resources.
At a glance
- What is it?
- carrot is a curated list of free ChatGPT and AI sites, published both as a README and as a separate website, with markers for free, recommended, recently updated, login-required and international-network entries. The list is maintained through issues, and its two tables overlap for eight of twelve rows.
- Who is it for?
- carrot fits someone looking for a reachable AI assistant and willing to read twelve rows to work out which ones are free. It does not fit anything that needs a filtered and verified directory, because the legend defines five markers that the tables never use as columns, the trending table duplicates the category table, and the descriptions are frequently truncated mid-sentence.
- Can I use it commercially?
- Not without permission. GitHub finds no licence file in the repository, and without a licence all rights are reserved by default: you may read the code but not reuse it. Check the README, or ask the authors, before using it.
- Is it still maintained?
- Yes. The repository last received commits 1 day ago.
- What is it written in?
- GitHub does not report a main language for this repository.
Answers come from the project's GitHub data, last synced on November 15, 2026, and from our analysis. They are not legal advice.
Star history
- xx025/carrot17,224 ★ · +34
| Repository | 2026-09-14 | 2026-11-15 |
|---|---|---|
| xx025/carrot | 17.2k | 17.2k |
34 Stars in the last 62 days, measured from the daily GitHub snapshot. Compare with other projects
Editorial analysis
The repository is one README, and the website is the product
Look at the top level of the repository and there are two entries: a .github directory and README.md. That is the whole project. There is no licence file, the primary language is not recorded, there are no releases, and the last push recorded for the repository is 2026-09-05. What the README does have is a published website, cc.ai55.cc, which it asks readers to bookmark and share, and the description says the project is a directory of AI tools to help you quickly find free, practical and efficient website resources. So the thing you consume is a web page, and the repository is a source for one page of it plus a place to send corrections. That distinction matters when you judge the project: you are reading a feed, not an application, and there is no build here to run.
The legend defines five markers and the tables use them as loose text
The legend table at the top of the file defines five markers: recently updated, recommended, free to use, requires a login or password, and needs international network access. That is a good schema for a directory of this kind, because each of those is a question a reader actually has. The problem is in the entries. Each row is a table row with a rank, a name linking out, a one-line description and a second link, and the cells that would hold the recommended and free markers are empty. The markers that do appear are typed into the start of the description text instead, so a reader looking for the free ones has to read every row. A legend that is not a column is a convention, and a convention does not survive contact with a table that someone edits by hand.
The trending table is a prefix of the category table
The first section is a list of what is hot, twelve rows, and the second is general assistants and smart search. Read them side by side and the first eight rows of the second table are identical to the first eight of the first, down to the descriptions and the link cells, and then the second table continues with entries the first does not have, Kimi among them. The practical consequence is that the two sections are not two opinions. The hot list is a copy that stops early, and the category list is the longer version. It also means the rank numbers are not comparable between them, since a row numbered 1 in the first table is the same site as the row numbered 1 in the second, and there is no way to tell from the page whether a site is genuinely trending or simply came first.
Descriptions stop mid-sentence with an ellipsis
The one-line description is the only editorial content in the list, and several of them are cut off. The entry for Claude ends after saying it is Anthropic's general assistant, good at reasoning, writing and analysis, with an ellipsis where the rest of the sentence would be. The entry for Qianwen ends after question answering, creation, reasoning and multimodal interaction, again with an ellipsis. This is a common way to keep a table narrow, and it is the wrong thing to do in a directory. The truncated part is usually the part that would distinguish the service from its neighbours, so the entries that need the sentence most are the ones losing it, and a reader has to open the site to learn what the list was going to tell them.
The most useful filter is a marker, not a column
Of the five markers, one is more valuable than the rest for a large group of readers: the one for entries that need international network access. Reachability is the first question when you are choosing an assistant from outside that network, and it is the question a curated list is uniquely placed to answer, because it is not in the product's own marketing. It is in the legend, but no table has a column for it. The other markers describe the service; this one describes whether you can reach it at all. It is the one worth turning into a real column, and it is the one a reader has to infer by reading prose.
Corrections arrive through issues, and that is the whole process
The file states its maintenance model in one sentence: submit sites, report entries that have stopped working, or correct information, through the repository's issues. That is a reasonable model for a list like this, and it has one virtue worth naming, which is that anyone can fix a dead entry without waiting for a release. The opening of the file makes no statement about verification, and the list is a set of third-party services reached by following a link, some of which, by the legend's own admission, need a login or a password. So the accuracy of any given row depends on somebody noticing, which is exactly the job the issues process is set up to do. If you are about to spend an afternoon testing assistants, this is a starting list of candidates, not a place to take a recommendation from.
A list of free services is a list of links nobody has checked for you
The honest way to use a directory like this is to treat it as a set of pointers and to follow each one to its own domain. Every entry is an external site, reached from a page that carries no licence, no review process and no security posture, and some may ask for a login or a password, which the legend flags as its own marker. That combination is not a warning about the listed services, most of which are large and well known. It is a warning about the path. A domain you learned from a listicle is a domain you did not choose, so type the address yourself rather than clicking through, and never enter a password on a domain a link list introduced you to. If a row is wrong, the fix is an issue, and if a row is malicious, the list has no mechanism to tell you which.
Editorial conclusion
carrot fits someone looking for a reachable AI assistant and willing to read twelve rows to work out which ones are free. It does not fit anything that needs a filtered and verified directory, because the legend defines five markers that the tables never use as columns, the trending table duplicates the category table, and the descriptions are frequently truncated mid-sentence. Before you rely on it, open the services from their own domains rather than through the list, treat the free marker as an unverified claim, and read the network-access marker for your own connectivity. And if you want to correct an entry, the project asks for an issue, since the repository is a single README file with no releases and a last recorded push on 2026-09-05.
Frequently asked questions
What is the carrot list?
It is a curated list of available AI assistants, agents, developer tools, creative tools and industry applications, presented as a free ChatGPT site list. It is published both as a README in the repository and as a separate website that the file asks readers to bookmark.
How do I submit a site or report a dead link in carrot?
The project asks for submissions, reports of entries that have stopped working, and corrections to go through the repository's issues. The repository itself is a single README file, with no releases and a last recorded push on 2026-09-05.
Does carrot mark which sites are free to use?
The legend defines a marker for free use, along with markers for recently updated, recommended, requiring a login or password, and needing international network access. In the tables the dedicated cells for those markers are largely empty, and the markers that do appear are typed into the description text, so the list cannot be filtered by them.
Which AI sites are in the carrot popular list?
The visible portion of the list names ChatGPT, Gemini, Claude, DeepSeek, Doubao, Qianwen, Perplexity, ChatGLM, Manus, Cursor, Midjourney and Hailuo video, each with a short description and a link. The general assistants section repeats the first eight of those and then adds others, including Kimi.
Official sources
Add this badge to your README
If you maintain this project, the badge below links readers to this analysis and shows its maintenance status from the daily GitHub snapshot. Paste the markdown into your README; add ?metric=license or ?metric=stars to the image URL for a different field.
[](https://hysenlabs.com/projects/xx025-carrot)Star history chart for your README
[](https://hysenlabs.com/projects/xx025-carrot)Was this useful?