JetBrains 2026: 90% Developers Use AI Coding Agents, Claude Code Surges to 39%

JetBrains 2026 AI Coding Agent Adoption Survey

JetBrains surveyed over 15,000 professional developers worldwide in early 2026. The results show that 90% of respondents use a coding agent at least weekly , and 68% use one daily , indicating AI-assisted coding has become a mainstream practice.

Tool Adoption Trends (January vs. May–July 2026)

Claude Code : Global work adoption rose from 18% to 39%; in the U.S. sample, adoption reached 47%, with 31% naming it their most-used tool.

GitHub Copilot : 21% work adoption (May–July).

Codex : Work adoption jumped from 3% to 16%; awareness grew from 27% to 65%.

Cursor : Awareness increased from 69% to 75%, but work adoption fell from 18% to 12%.

OpenCode (open source) : 7% work adoption, 42% awareness.

From Tool Comparison to Task Completion

The competition is no longer about which editor is better, but which agent can complete a task end-to-end with the least friction . JetBrains IDEs now integrate Claude Agent, Codex, Copilot, OpenCode, and others via the ACP (Agent Client Protocol), enabling multiple agents to run simultaneously—one writing code, another writing tests, another doing reviews.

Code Generation Speed ≠ Delivery Speed

The article illustrates with a concrete example: asking an agent to add a payment callback endpoint (controller, signature verification, order update, logging). The agent may produce the code in minutes, but the developer cannot ship it without verification:

Will duplicate callbacks charge twice? Are there signature validation gaps? What happens if the database write fails? Will high concurrency cause issues?

These concerns require business knowledge, system understanding, and awareness of failure modes—skills that go beyond syntax generation. The author emphasizes: "Code generates fast, but delivery doesn't necessarily speed up. Generating 1,000 lines takes minutes; review, integration testing, supplementing tests, and production debugging can still take half a day."

Accountability Remains with the Developer

"AI writes the code; we still carry the can." The core value of a programmer shifts from typing speed to:

Understanding what problem needs solving

Knowing how to solve it correctly

Being able to verify, modify, and validate AI output

Taking responsibility for the result in production

Using agents is becoming a baseline skill; the differentiator is the ability to judge correctness, decide release readiness, and own the outcome.

Reference

Original JetBrains report:

https://blog.jetbrains.com/research/2026/08/ai-coding-agent-adoption-2026/