AI coding agents have become a regular part of developers’ toolkits, according to the results of the tenth edition of the Developer Ecosystem Survey 2026, conducted by JetBrains among more than 15,000 professional developers worldwide. The survey indicates that 90% of respondents used coding agents at work at least once a week between May and July 2026, while 68% used them daily, whether through local agents or remote cloud services.
These results indicate that coding agents have shifted from a recent experiment to a practical component of daily workflows. However, the usage figures in the report reflect respondents’ answers to the survey and do not necessarily mean that all uses are equal in terms of depth or reliance within development teams.
Claude Code Leads the Adoption Race
Claude Code continued to grow at a notable pace, becoming the most widely adopted AI coding tool at work according to the survey. Between May and July 2026, approximately 39% of professional developers worldwide used it, up from 18% in January 2026. In the United States, the figure reached 47%, meaning the tool was used by nearly half of the participating developers from that country.
Claude Code’s presence is not limited to occasional use; 31% of developers said it was their most-used AI coding tool. According to JetBrains, this represents a conversion rate of approximately 80% from regular use to considering the tool the primary one within users’ AI toolsets.
Rapid Growth for Codex and a Decline in Copilot’s Position
Codex recorded a major jump, with adoption rising from 3% in January 2026 to 16% between May and July, nearly a fivefold increase. The percentage of developers who had heard of the tool also rose from 27% to 65% during the same period.
By contrast, GitHub Copilot lost the leading position it had held for a long time. Its adoption declined from 29% a year earlier to 21% in May–July 2026, but it remains one of the tools with the strongest developer awareness, at 79%. This figure rises to between 86% and 90% in Europe, the United Kingdom, and the United States. The survey also found that 39% of GitHub Copilot users use it, among other environments, inside JetBrains integrated development environments.
Contrasting Performance for Cursor and Open-Source Tools
Developer awareness of Cursor rose from 69% in January to 75% in May–July, but its adoption declined from 18% to 12%. The sharpest decrease occurred in China, where usage fell from 28% to 16% during the same period.
OpenCode, an open-source coding agent, reached an adoption rate of 7%, with global awareness at 42%, despite not being associated with a major company, according to the report’s description.
Adoption of Google Antigravity remained at 6%, alongside a significant increase in awareness from 29% to 47%. India remains its strongest market in the survey data: 15% of developers there use it, up from 10% in January, bringing it close to Cursor in third place among the most-used tools in India.
What Does This Mean for Developers’ Workflows?
The data shows that competition is not only about the number of users, but also about an tool becoming the primary choice within a developer’s toolset. The increase in awareness of tools such as Codex and Google Antigravity, alongside differing adoption rates, also shows that product awareness does not automatically lead to regular use.
JetBrains is likewise working to expand its position within this ecosystem. According to the article, approximately 9% of developers worldwide use JetBrains AI inside integrated development environments and/or Junie at work. Claude Agent, Codex, GitHub Copilot, and OpenCode also integrate directly with AI chat in JetBrains environments, and dozens of other agents, including Cursor, can be added through the Agent Client Protocol (ACP). JetBrains also enables users to use Codex through an OpenAI API key or a ChatGPT subscription.
The company is currently testing Air, an agent-based development environment available in preview, with the goal of combining multiple agents in a single workflow. It also offers JetBrains Central as a unified control and execution layer for agent-based development processes across different tools and environments.
Data Limitations and Survey Methodology
The report states that the definition of “professional developers” included participants who work in programming or development in roles such as developer or software engineer, AI or machine learning engineer, DevOps or infrastructure engineer, architect, data scientist or engineer, and quality assurance engineer. Developers, programmers, and software engineers made up approximately 90% of the sample.
The survey was localized into eight languages, with regional quotas established to represent the global distribution of developers, after which the results were statistically reweighted according to region, employment status, programming language, and familiarity with JetBrains products. Therefore, comparisons should be read as indicators of adoption trends in the reweighted sample, not as a comprehensive record of all developers or a final judgment on tool quality.