Claude Code Dethrones Copilot: JetBrains’ 2026 AI Coding Survey

JetBrains' August 2026 AI Coding Agent Adoption report, based on 15,000+ professional developers, shows Claude Code now commands 39% adoption worldwide and 47% in the U.S., used twice as often as GitHub Copilot, which slid from 29% to 21% in a year. OpenAI's Codex surged from 3% to 16% adoption in six months, while Cursor dropped from 18% to 12% amid the pending SpaceX-Anysphere acquisition. The report signals a structural shift from inline suggestion engines to agent-first workflows, with 90% of developers using AI tools weekly and 68% daily.

JetBrains just published its August 2026 AI Coding Agent Adoption report — 15,000+ professional developers, tenth year running — and the numbers tell a story GitHub would rather ignore. Claude Code now commands 39% adoption worldwide, reaches 47% in the United States, and is used twice as often as GitHub Copilot, which has quietly slid from 29% to 21% in a single year. That reshuffling happened in roughly 18 months. How Claude Code Got Here Eighteen months ago Claude Code had 3% adoption. By January 2026 it was at 18%. By mid-2026 it hit 39% — and 31% of all surveyed developers now call it their primary AI coding tool. The conversion rate is striking: of developers who use Claude Code regularly, 80% make it their default. Its 91% CSAT score and NPS of 54 are numbers most SaaS companies would frame on a wall. The drivers are not mysterious. Claude Code was built for terminal-native, agent-first workflows from the start — not retrofitted. Access to Opus 5 the current performance leader at 1691 Elo and genuine computer-use capabilities gave it a ceiling no competitor currently matches. In JetBrains’ blind preference tests, 67% of developers preferred it over alternatives without knowing which tool they were using. What Actually Happened to Copilot GitHub Copilot’s defenders will point to its 79% awareness score and call this a blip. That defense doesn’t hold. Awareness is a lagging indicator — it reflects where the market was , not where it’s going. Developers know what Copilot is; they’re choosing something else. Several things broke trust simultaneously. Suggestion quality degraded; acceptance rates sit at 35–40% compared to competitors running 42–45%. In March 2026, Copilot injected promotional content into over 1.5 million pull requests https://www.nxcode.io/resources/news/github-copilot-getting-worse-2026-developers-switching , triggering a wave of complaints and tool-switching. Then Microsoft restructured pricing into a credit pool model where heavy agent usage generated variable, unpredictable bills. Each of these alone would have been manageable. Together, they accelerated an exit. Copilot’s remaining moat is enterprise procurement. In companies with 10,000+ employees, it still leads at 56% adoption, kept alive by contracts signed before the problems piled up. That moat is real — but it’s also finite. The Dark Horse: Codex OpenAI’s Codex went from 3% adoption in January 2026 to 16% by mid-year — a 5x increase in six months. Awareness jumped from 27% to 65% in the same window. If that trajectory holds, Codex will be a serious 2 contender before year-end. The tool launched with an explicitly agentic workflow that resonated with developers who want less friction, not more configuration. Cursor, once the darling of the AI-native IDE wave, is sliding. Adoption dropped from 18% to 12% even as awareness grew. The pending SpaceX-Anysphere acquisition is generating legitimate uncertainty — developers are reluctant to deepen dependence on a tool whose ownership is in flux. China saw the sharpest drop: 28% to 16%. The Structural Shift The JetBrains survey https://blog.jetbrains.com/research/2026/08/ai-coding-agent-adoption-2026/ confirms what the ecosystem has been signaling for months: the inline suggestion engine era is over. Tools architected around autonomous, multi-step agent workflows are pulling ahead of tools that bolted agent capabilities onto an autocomplete core. Claude Code and Codex were built for the new paradigm. Copilot and Cursor were built for the old one and are retrofitting — and it shows. AI coding adoption is now essentially universal among professional developers. 90% use these tools at least weekly; 68% use them daily. The question has shifted from whether to use AI tools to which ones — and that’s a more competitive race than any vendor anticipated. What This Means for Your Stack If you haven’t re-evaluated your current AI toolchain against this data, now is the time. A few practical observations: - Claude Code’s 80% conversion rate suggests it earns its keep on merit — give it a real evaluation, not a 20-minute demo. - Codex at 5x growth in six months deserves a serious look, especially if you’re already in the OpenAI ecosystem. - If you’re on Copilot through enterprise licensing, check what you’re getting versus the credit ceiling. The billing model changed. - OpenCode 7% adoption, 194k GitHub stars, MIT licensed is the strongest choice for teams with air-gap, sovereignty, or vendor-lock concerns. The best-of-breed era has arrived. According to JetBrains’ earlier January survey https://blog.jetbrains.com/research/2026/04/which-ai-coding-tools-do-developers-actually-use-at-work/ , developer satisfaction is driving tool selection faster than enterprise procurement cycles anticipated — and that trend has only accelerated. Copilot had three years of near-monopoly. The market is making up for lost time, and every major ranking now reflects a reshuffled top three https://blog.logrocket.com/ai-dev-tool-power-rankings/ .