JetBrains Developer Ecosystem 2026: What 15,509 Developers Say About AI Coding Agents

JetBrains has published a snapshot from its tenth annual Developer Ecosystem Survey. The survey ran from May through July 2026 and the analysis covered 15,509 professional developers. It was available in eight languages and used regional quotas plus statistical weighting for region, employment status, programming languages, and familiarity with JetBrains products.

The main results:

• 90% of respondents use AI coding agents at work at least weekly, and 68% use them daily;
• Claude Code rose from 18% in January to 39%. It reached 47% in the United States, while 31% of developers named it their primary AI tool;
• GitHub Copilot declined from 29% to 21%, and Cursor from 18% to 12%;
• Codex rose from 3% to 16%, while awareness increased from 27% to 65%;
• OpenCode reached 7%, Google Antigravity 6%, and JetBrains AI Assistant and/or Junie 9% combined.

The growth of Claude Code and Codex is a strong market signal: the mix of workplace tools is changing quickly. The word adoption, however, is easy to read too broadly.

The headline speaks about agents, but the chart asks a broader question: which AI tools respondents use for work. Agents, assistants, and editors appear in the same list; respondents could select multiple answers, and the chart shows only a subset. The 39% and 21% figures are therefore not market shares, counts of paid seats, or telemetry for daily activity. One developer could select Claude Code, Copilot, and Codex at the same time.

Representativeness comes with another caveat. This is a large international sample adjusted through quotas and weights, but the published materials do not describe probability-based random selection. Instead, the survey used open opt-in recruitment: JetBrains distributed a link and offered prizes, a license discount, and rewards for referrals. A large N reduces random noise, but it does not remove self-selection — including elevated interest in AI or in JetBrains itself.

The detailed methodology link in the new article still points to the 2025 survey methodology. For the 2026 wave, the full recruitment mix, participation rate, country-level sample sizes, effective sample size after weighting, and raw data have not been published. The 47% US figure should therefore be read more cautiously than the global estimate.

There is one more limitation: the January figures come from AI Pulse, while the summer figures come from the Developer Ecosystem Survey. These are different cross-sectional samples, not observations of the same people. We can say that Copilot's reported share declined between surveys. We cannot say that specific users “switched” to Claude Code or establish why the change happened.

I would therefore use this study as market radar rather than a precision instrument. Claude Code has moved ahead among the tools shown, Codex has gained ground quickly, and the previous leaders have declined. A company making its own decision still needs its own instrument panel: active use, accepted changes, review time, rework, incidents, and cost per useful task. Otherwise, the global 90% becomes another attractive adoption number without an answer to what actually improved.

#AI4SDLC #AI #Agents #Research #Metrics #DevTools