Which AI coding tools do developers use? Takeaways from the JetBrains 2026 survey
Workplace usage rates of Claude Code, Copilot, Codex and Cursor according to a survey of more than 15,000 developers.
In the software world, the question “do you use AI?” has given way to “which one do you use?”. JetBrains’ Developer Ecosystem Survey 2026 (the tenth year of the survey), shared in August 2026, offers an up-to-date snapshot. More than 15,000 professional developers worldwide took part; usage data was collected between May and July 2026.
First, the big picture
- 90% of professional developers use AI coding agents at work at least once a week.
- 68% use them every day.
So these tools are no longer a curiosity or an experiment; they are part of the daily workflow.
Professional developers
How much is each tool used?
Workplace usage rates:
| Tool | Usage | Change |
|---|---|---|
| Claude Code | ~39% | 18% in January 2026 |
| GitHub Copilot | 21% | 29% a year ago |
| Codex | 16% | 3% in January 2026 |
| Cursor | 12% | 18% in January 2026 |
| JetBrains AI (IDEs and/or Junie) | ~9% | n/a |
| OpenCode | 7% | n/a |
| Google Antigravity | 6% | Flat |
JetBrains Developer Ecosystem Survey 2026, percentage of developers
Source: JetBrains Blog, AI Coding Agents: Adoption Trends (August 2026). The survey was run on JetBrains' own audience.
Standout stories
Claude Code rose to the top quickly. According to the survey results it became the most-used tool; in the US the rate reaches 47% and 31% of developers name it as their most-used tool. The report says it is used roughly twice as often as Copilot.
Copilot lost the lead but is highly recognized. Awareness is 79%, reaching 86–90% in Europe, the UK and the US. 39% of its users use Copilot in JetBrains IDEs.
Codex is catching up fast. Usage grew about fivefold; awareness rose from 27% to 65%. Much of the growth comes from awareness.
Cursor’s story is the reverse. Its awareness rose from 69% to 75% while usage fell. The biggest drop is in China: from 28% to 16%.
How should you look at this data?
This survey was run on JetBrains’ own audience, meaning participants are mostly developers who know and use JetBrains products. The real-world distribution may differ somewhat. Still, because it has been run with the same method for ten years, it is valuable for showing trends (who is rising, who is falling).
What does it mean for your own projects?
For small teams or solo projects, the takeaway is this: when choosing a tool, pick the one that fits your workflow, not the “most popular” one. Agents running in the terminal, assistants embedded in the editor and agents running in the cloud appeal to different habits. Re-trying them once a month and comparing against your own yardstick works better than being a fan of a single tool.