JetBrains just dropped the Developer Ecosystem Survey 2026 results and the AI coding agent market has completely flipped. Claude Code is now the most widely adopted AI coding tool at work — used by 39% of professional developers globally, double GitHub Copilot’s 21%.
For anyone tracking this space, the numbers confirm what we’ve been seeing anecdotally. But the velocity of the shift is staggering.
The Headlines
- Claude Code: 39% adoption (up from 18% in January 2026) — now the #1 AI coding tool
- GitHub Copilot: 21% adoption (down from 29% a year ago) — lost the lead it held since 2023
- Codex: 16% adoption (up from 3% in January) — 5x growth in six months
- Cursor: 12% adoption (down from 18% in January) — declining globally
- OpenCode: 7% adoption — impressive for an open-source tool with no big company behind it
The survey ran May–July 2026, covering 15,000 professional developers worldwide. This isn’t a Twitter poll or a vendor-commissioned white paper — it’s statistically reweighted to represent the global developer population.
Claude Code’s Conversion Rate Is the Real Story
Raw adoption tells part of the story. Conversion tells the rest.
31% of developers say Claude Code is their most used AI coding tool — which means an almost 80% conversion rate from “I use it at work” to “it’s my primary tool.” That’s stickiness. Claude Code isn’t just being tried; it’s becoming the default.
In the US, adoption is even higher: 47% of American developers use Claude Code at work. Nearly half the market in eight months.
Copilot Lost the Lead It Invented
GitHub Copilot created the AI coding assistant category in 2023. It was the long-standing leader. That era is over.
Copilot dropped from 29% adoption a year ago to 21% today. It still has 79% mindshare (awareness), but awareness without conversion is a warning sign. Developers know it exists; they’re just choosing something else.
The survey notes that 39% of Copilot users still use it in JetBrains IDEs — so it’s not abandoned, just no longer dominant.
Codex’s 5x Growth in Six Months
OpenAI’s Codex went from 3% to 16% adoption in six months. That’s the fastest growth rate in the survey.
More striking: awareness jumped from 27% to 65%. Six months ago, most developers hadn’t even heard of Codex despite the OpenAI brand. Now two-thirds know it, and a quarter of those who know it are using it.
The ChatGPT integration is clearly helping. Codex benefits from being the coding agent inside a product developers already use daily.
Cursor’s Decline
Cursor dropped from 18% adoption in January to 12% in May–July. The biggest decline happened in China — 28% to 16% in the same period.
Mindshare is still strong (75%, up from 69%), so developers know the product. But conversion from awareness to active use is slipping.
Why? Part of it may be market competition — Claude Code and Codex both improved significantly this year. Part of it may be regional dynamics. The China decline is worth watching.
The 90% Number That Matters
90% of professional developers now use AI coding agents at work at least weekly. 68% use them daily.
This isn’t a niche tool category anymore. It’s infrastructure. If you’re building software without an AI agent in 2026, you’re the outlier.
OpenCode’s 7% With No Marketing Budget
OpenCode — the open-source coding agent — reached 7% adoption with 42% mindshare. No big company behind it. No Super Bowl ad. Just word of mouth and GitHub stars.
That’s the highest organic penetration in the survey. It proves developers will adopt tools that respect their autonomy and don’t lock them into a vendor ecosystem.
Google Antigravity’s India Stronghold
Google Antigravity is stable at 6% adoption globally, but India tells a different story — 15% adoption, tied with Cursor as the #3 AI coding tool in that market.
India is becoming a bellwether for AI coding tool adoption. The market is price-sensitive, mobile-first, and willing to experiment. What wins there often signals where the global market goes next.
What This Means
Three takeaways for anyone building, buying, or betting on AI coding tools:
-
The leader changed in eight months. Claude Code went from 18% to 39%. Copilot went from the default to second place. In a market this young, nothing is locked in.
-
Model quality alone doesn’t explain it. Claude Code’s stickiness (80% conversion) suggests the harness matters as much as the model. Developers aren’t just picking the smartest model — they’re picking the agent that fits their workflow.
-
Open source can win. OpenCode’s 7% with no corporate backing proves you don’t need a trillion-dollar company to reach developers. You need a tool that works.
The full survey methodology is worth reading — 15,000 respondents, eight languages, regionally reweighted. This is the best data we have on who uses what.
Source
- AI Coding Agents: Adoption Trends — JetBrains Research (August 18, 2026)