Claude Code vs GitHub Copilot vs Cursor: What JetBrains' 15,000 Developer Survey Actually Shows

JetBrains Research's 2026 Developer Ecosystem Survey is one of the largest independent looks at how professional developers actually use AI coding tools day to day, with fieldwork running May through July across more than 15,000 respondents, roughly 90 percent of whom identified as developers, programmers, or software engineers, with the rest spread across AI and ML, DevOps, architecture, data, and QA roles. The topline finding getting attention is simple: Claude Code more than doubled its workplace adoption in six months while both GitHub Copilot and Cursor lost ground over the same period.

The headline shift, and why the direction matters more than the exact number

Claude Code's reported adoption climbed from about 18 percent in January 2026 to 39 percent by the May through July fielding window, with US adoption specifically at 47 percent. Thirty one percent of respondents named it their primary or most used AI coding tool, which is a meaningfully stricter bar than simple usage, since a developer can use several tools without committing to one as primary. Over the same stretch, GitHub Copilot fell from roughly 29 to 31 percent down to 21 percent, and Cursor fell from about 18 percent to 12 percent. OpenAI Codex moved the opposite direction entirely, climbing from about 3 percent to 16 percent.

Point estimates from any single survey wave carry sampling noise, and this is not a controlled longitudinal panel following the same developers over time. What makes the pattern worth taking seriously anyway is that four separate tools all moved by double digit percentage points in the same six month window, in a consistent direction for each: Claude Code and Codex up sharply, Copilot and Cursor down. A single outlier swing is easy to dismiss as noise. Four tools moving together, in a sample this large, looks more like a real reallocation of developer attention than a measurement artifact.

What "named as primary tool" actually isolates

The 31 percent figure for Claude Code as a named primary tool matters more than the 39 percent general adoption number for teams making a standardization decision, because general adoption includes developers trying a tool occasionally or keeping it installed without relying on it for real work. A tool becoming someone's default, the one they reach for without thinking about alternatives, is a stronger signal of actual day to day value than simple installed base. GitHub Copilot's awareness sits at 79 percent, identical to Claude Code's, which tells you the two tools are equally well known among developers; the divergence in adoption after that point is not a discovery problem for Copilot, it is a retention and preference problem.

The caveat worth taking seriously before you act on this

The survey results reached wider audiences through 36Kr's reporting, and it is worth noting plainly that 36Kr's status as a primary source for the underlying JetBrains data has not been independently established in every account of this survey, and the available reporting does not specify whether respondents could select multiple tools or how the sample was weighted across company size, region, or seniority. None of that invalidates the directional finding, a large, recent, broad sample showing Claude Code's adoption more than doubling is a real data point, but it is a reason to treat the exact percentage points as approximate rather than precise, and to look for your own organization's usage data before making a tooling decision off a single external survey alone.

What this means for a team standardizing its toolchain

A separate, related JetBrains figure is arguably more important than any single tool's market share: about 90 percent of respondents reported using an AI programming agent at work at least once a week, and 68 percent reported daily use. The real decision most engineering leaders face now is not whether to adopt an AI coding agent, that ship has sailed for the large majority of the industry, it is which one to standardize on, how many tools to support in parallel, and how to budget for a category where usage patterns are shifting by double digits every few months. Teams evaluating Claude Code specifically against Cursor's IDE centric workflow should weigh this adoption data alongside their own team's actual workflow fit, since a tool with rising industry wide adoption is not automatically the right fit for every codebase and team structure.

When this applies to your stack

If your organization is still running an informal mix of whatever individual developers picked up on their own, this survey is a reasonable prompt to formalize that decision rather than let it drift further, especially given how much adoption moved in just six months. Choosing a standard agentic coding tool, and the licensing, security review, and workflow integration that comes with that choice, is exactly the kind of developer tooling and agent engineering decision we help engineering organizations work through before the fragmentation gets expensive.

FAQ

Which AI coding tool had the biggest adoption gain in JetBrains' 2026 survey? Claude Code, rising from about 18 percent workplace adoption in January 2026 to 39 percent by the May through July fielding window, more than doubling over six months.

Did GitHub Copilot lose developers to Claude Code specifically? The survey does not establish direct switching between any two tools. It shows Copilot's adoption fell from roughly 29 to 31 percent down to 21 percent over the same period Claude Code rose, which is suggestive but not a causal claim the data itself makes.

How reliable is this survey's methodology? JetBrains Research surveyed more than 15,000 professional developers, a large sample, but widely circulated reporting on the results does not specify whether respondents could select multiple tools or how the sample was weighted, so treat exact percentages as directional rather than precise.

What percentage of developers use any AI coding agent at work? About 90 percent reported using one at least weekly, and 68 percent reported daily use, which is a stronger signal of category wide adoption than any single tool's market share number.