Claude Code ranks second in terms of the number of AI Coding Agents used. What's number 1? ~Summary of AI Coding trends in July 2025 as seen through data~
AI Coding Agent July trends
We have released a website called AI Coding.Info.
This is a site that observes usage trends related to AI Coding Agents such as Claude Code, Gemini, or Codex from a fixed point from information in the Github repository. To determine the use of AI Coding Agent, we conduct daily surveys under the following conditions.
- Top 100 public repositories with the most stars on Github for 30 programming languages
- 16 types of AI Coding Agents
- Only when the rule files used by various AI Coding Agents are in the Github repository, it is determined that AI Coding Agent is being used.
AI Coding Agent usage rate is 2.9%
**AI Coding Agent usage rate is 2.9%. ** This is the percentage of repositories that showed evidence of AI Coding Agent usage, based on a survey of 3,000 repositories. This can be said to be a small percentage of the total.
https://ai-coding.info/ja/agents
This is due to multiple factors. One is that this survey is based on Github's top 100 stars. Therefore, can it be said that the usage status of Github's top 100 stars represents the usage status of AI Coding Agents around the world? There is a problem. Furthermore, since this is based on a public repository, there may be a discrepancy with the current usage situation of companies. Rather, it represents the adoption status of OSS. You can also say that. Also, as a mechanical problem, if rule files such as CLAUDE.md and GEMINI.md do not exist, it is not possible to determine whether they are being used or not. There is also the problem. Therefore, a repository that uses Claude Code but does not create a CLAUDE.md file will be judged as not using Claude Code.
Share of AI Coding Agent by product
**The most used product is "Cursor", followed by "Claude Code", and third is "Copilot Agent" (Github Copilot). ** Considering that Gemini CLI was announced on June 25th, the fact that it has been in 4th place for almost a month can be said to be quite widespread. However, on the other hand, it can be said that Gemini CLI has only about 1/3 of the available repositories of Cursor.
https://ai-coding.info/ja/agents
AI Coding Agent usage status by programming language
The programming language most used by AI Coding Agent is "Typescript", followed by "Python" and "Rust". As for TypeScript, AI Coding Agent seems to have originated from VS Code extensions such as Github Copilot, so if you think about it from that perspective, I think this is a somewhat understandable result. Also, considering that Python has a high affinity with ML systems such as DeepLearning, this result may not be so strange. From there, the third Rust was a surprising result. If you imagine the result of this, Rust is a programming language that has been gaining popularity over the past few years. There is a recognition that This may be because the programming language community is young and eager to learn about new initiatives like AI Coding Agent. You might say that.
https://ai-coding.info/ja/languages
Monthly trends in the number of repositories used by AI Coding Agent
Let's take a look at the trends in the number of repositories (with overlap) using AI Coding Agent in July 2025. **As of July 1st, we have confirmed the use of AI Coding in 77 repositories. As of July 31, the number had increased to 112 repositories. The number of repositories in use has increased by 1.4 times. **
In the past month since we launched the service, we have added AI Coding Agents such as Trae IDE and Junie, and changed the usage standards for Gemini, and we can see that AI Coding Agents are permeating development at a fairly rapid rate.
Difference in usage of AI Coding Agent in programming languages
The survey revealed that there are large differences in the usage rate of AI Coding Agent depending on the programming language. **For TypeScript, AI Coding Agent is used in 21% of all repositories, while for Go language it is only about 5% of the total. **
https://ai-coding.info/ja/languages/typescript
https://ai-coding.info/ja/languages/go
Although there is no solid evidence regarding this, I heard some interesting opinions during interviews about the service. The content was that "coding using generation AI is prohibited in infrastructure-related OSS." There are several examples of this when collecting information.
https://www.netbsd.org/developers/commit-guidelines.html
https://wiki.gentoo.org/wiki/Project:Council/AI_policy
https://github.com/qemu/qemu/commit/3d40db0efc22520fa6c399cf73960dced423b048
These are NetBSD,'' GentooLinux,'' and QEMU,'' which are not OSS implemented using the Go language. However, in areas close to infrastructure such as OS and virtual machines, the use of generated AI is partially prohibited. Kubernetes is a famous OSS for the Go language. However, there is no evidence that generative AI is used in the surrounding ecosystem (CNCF related). In such infrastructure, mission-critical, and high-performance areas, it may still be prudent to use generative AI. This is paradoxical, but it may be said that Since the Go language is used in mission-critical high-performance areas, the generated AI code cannot be used because the risk of failure is too large.Therefore, Go language is rarely used (especially in famous OSS).'' If you look at the repositories that actually use Go, it seems like there are very few infrastructure-related ones. Is it about grafana/loki or cockroachdb?
https://ai-coding.info/ja/languages/go
To summarize what has been said so far, **AI Coding Agent seems to be popular, but its actual usage tends to be low at 2.9% of the total. However, on the other hand, the percentage of users using specific programming languages such as TypeScript is rapidly increasing. ** It can be said that This is what the other perspective I mentioned earlier is about.
What programming languages are compatible with AI Coding?
In the previous chapter, I talked about the difference between programming languages that can utilize AI Coding and those that cannot. I made a simple hypothesis here.
“Programming languages with many users probably have many repositories using AI Coding.”
That's what it means. So let's analyze the data. Let's plot the number of pushes for a specific programming language on Github and the number of repositories used by AI Coding Agent. Data regarding the number of pushes by programming language on Github is published in the repository.
https://github.com/github/innovationgraph/tree/main
This is the first hypothesis I made at the beginning.
“Programming languages with many users probably have many repositories using AI Coding.”
It seems that this hypothesis can be rejected. There was a negative correlation with the number of accounts pushed to Gthub in 2024. The above hypothesis was intended to show that ``Actually, the usage rate of AI Coding users is about 0% regardless of programming language.However, there are differences in the number of repositories used depending on the number of programming language users.'' But that's not the case. I understand that. Now, on the contrary
“Is AI Coding Agent often used for programming languages with a small number of users?”
It turns out that this is not the case. This is a programming language with fewer users than TypeScript, as divided into Group I and Group II in the graph, and there is no programming language that uses AI Coding Agent as many times as TypeScript. It becomes. Next, let's define the concept of "AI Coding Agent usage ratio." This is the value obtained by dividing the "number of repositories used by AI Coding Agent" by the "number of accounts pushed to Github in 2024." Plot this on the vertical axis and the "number of accounts pushed to Github in 2024" on the horizontal axis. Then, plot the "number of repositories used by AI Coding Agent" as the size of the circle.
There are two things that can be seen from this graph. One is that the number of repositories using Rust and Python AI Coding Agents is almost the same, but the breakdown is different. The usage rate of AI Coding Agent for Python is low. However, due to the large number of language users, there is a certain number of repositories in use. On the other hand, Rust has a very high usage rate of AI Coding Agent, but because there are few language users, the number of usage repositories remains at a certain level. Second, the usage rate of TypeScript's AI Coding Agent is not that high. TypeScript's usage rate is 4.65E-06, Go's usage rate is 5.00E-06, and Ruby's usage rate is 4.45E-06. Relatively speaking, it hasn't changed by 10%. In fact, there is not much difference in usage rate between TypeScript, Go, and Ruby. The difference in the number of repositories used may simply be due to a difference in the number of users. If these facts are correct, the number of Rust users may increase rapidly due to AI Coding Agent. ** This is due to the linguistic nature of Rust, so in principle it can be extended by AI Coding Agent. Rather, it is an inductive deduction based on observed facts, but Rust and AI Coding Agent seem to be compatible since the usage rate of Coding Agent is high. When you think about it, Rust is actually the language that can benefit the most from AI Coding Agent, and I have a feeling that it will grow significantly over the next year. However, currently, the number of AI Coding usage repositories by programming language that we have been able to investigate is 100, and of those, the number of AI Coding Agent usages is on the order of 20 at most, so the discussion could become quite sensitive if the number of usages fluctuates by just one or two.
thoughts
Many AI Coding Agents have been released. AI Coding.Info alone handles 16 types of products. My daily life is a little busy. If you delay catching up, the situation may quickly change. Additionally, even when trying to catch up on information, the sources were often biased, making it difficult to obtain accurate information. A while ago, I was using Cline. There was a lot of talk about this, but in fact, the one that was popular in Japan was RooCode. However, as you can see from AI Coding.Info, RooCode has almost no adoption on Github. As for Cline, it's in the Top 100 repositories on Github, so it's only used in itself and about one other repository. Also, as a big topic, I feel like there's been a lot of talk about ClaudeCode around me lately. However, on the other hand, Cursor is often used in repositories. There were also biases depending on the country you were in and the natural language you usually used. I want a source of information that I can use while still being able to make a calm judgment within myself. That's why I started this site. Check the trends in AI Coding and wonder which tools are compatible with the programming language you are using. I wonder what kind of rule files are actually written in famous OSS? If you are interested in this, please take a look.