Nine out of 10 professional developers use AI coding agents at least once a week, a survey found. Workplace use of Claude Code rose sharply, overtaking the long-time leader GitHub Copilot to become the most-used AI coding tool.

On Aug. 24, online media outlet Gigazine reported that in JetBrains Research's "Developer Ecosystem Survey 2026," 90 percent of professional developers said they use AI coding agents at least once a week.

The survey was conducted from May to July among more than 15,000 professional developers worldwide. About 90 percent of respondents were developers and programmers or software engineers, and the survey covered eight language groups including English, Spanish, Chinese, Japanese, Korean, German, French and Portuguese.

Use of AI coding agents was frequent. Some 68 percent of respondents said they use AI coding tools for work every day. That suggests AI coding tools have become part of routine development work, rather than auxiliary technology developers use only for testing.

A notable change was the ranking by tool. Claude Code took the top spot in workplace use at 39 percent. GitHub Copilot, previously a leading AI coding tool, followed at 21 percent.

Claude Code's rise was steep. Its workplace use rate was only 3 percent in early 2025, but it jumped to 39 percent by mid-2026. That amounts to a roughly 13-fold increase in about 18 months.

By contrast, GitHub Copilot had held at around 29 to 31 percent for some time, but fell to 21 percent in this survey. It remained highly recognised, but trailed far behind Claude Code in workplace use.

Awareness of the two services was the same at 79 percent. Recently, awareness of other AI coding tools has also been rising quickly.

OpenAI's Codex is a representative case. Only 27 percent of developers said they knew about Codex in January 2026, but that rose to 65 percent in the May to July survey.

Actual use also increased quickly. Codex had a 0 percent workplace use rate in early 2025, but recorded 16 percent by mid-2026. JetBrains Research assessed Codex as a tool that is spreading rapidly.

The survey also found that the open-source-based OpenCode and Google Antigravity secured a certain level of awareness and adoption. JetBrains AI, which links multiple AI agents, was also used by some developers.

The changes shown in the survey indicate that developers' use of AI itself is shifting. Rather than tools that simply autocomplete code or support specific tasks, agent-style AI is moving toward participating in the overall flow of development work.

JetBrains Research also analysed the results as showing that agent-style software development could advance further in the future.

The survey also confirmed that high awareness in the AI coding market does not necessarily translate into high workplace use. Even with the same awareness, GitHub Copilot and Claude Code showed a large gap in actual adoption, and cases also emerged in which later tools such as Codex quickly increased users in a short period.

Ultimately, in future competition, a key advantage is expected to be how deeply a tool is embedded in actual development work, rather than which AI coding tool developers know. Attention is focused on how quickly AI coding agents will change developers' everyday ways of working.