Which AI Coding Tools Developers Use at Work

In January 2026, JetBrains surveyed more than 10,000 professional developers worldwide and found that 90% already use at least one AI tool at work for coding tasks. That is a huge number, but the more interesting detail is which tools actually made it into day-to-day work.

The answer is less about hype and more about habits. JetBrains’ AI Pulse survey shows that developers are splitting into two camps: chatbot users who ask for help in a browser tab, and tool users who want AI inside their editor, terminal, or agent workflow.

Most developers already use AI at work

The headline number is hard to ignore. By January 2026, 90% of developers said they regularly used at least one AI tool at work for coding and development tasks. JetBrains also found that 74% had adopted specialized AI tools for developers, meaning assistants, editors, and agents rather than general chatbots.

That matters because the market is no longer about whether developers will try AI. It is about which products earn a permanent place in the workflow. The tools that win are the ones that reduce friction inside the editor, terminal, or codebase, not the ones that merely answer questions well.

• 90% of developers regularly used at least one AI tool at work
• 74% had adopted specialized AI tools for developers
• The survey sample included more than 10,000 professional developers
• The research was localized into eight languages

JetBrains’ data also shows that developer usage is broad, not confined to early adopters or hobbyists. The survey covered developers, AI and ML engineers, DevOps engineers, architects, data scientists, and QA engineers who code. Roughly 90% of the sample fell into the developer or programmer category.

That makes the results useful for workplace reality, not just online chatter. If you are trying to understand what software teams actually buy, standardize on, or experiment with, this is closer to the truth than a product launch thread.

Claude Code and Copilot are winning real work

Claude Code is the fastest riser in the report. In January 2026, 57% of developers had heard of it, up from 49% in September 2025 and 31% in April to June 2025. Adoption at work reached 18%, which is roughly 1.5 times higher than in September 2025 and about 6 times higher than the 3% level seen in April to June 2025.

That growth is backed by strong satisfaction. JetBrains reports a CSAT score of 91% and an NPS of 54 for Claude Code, the highest loyalty metrics in the study. In plain English, people who try it tend to like it enough to recommend it.

"The shift toward best-of-breed agents demonstrates that product excellence now outweighs ecosystem lock-in."

GitHub Copilot is still the most widely known tool in the survey, with 76% awareness and 29% adoption at work. But its growth has flattened. It remains especially strong inside very large companies, where adoption reaches 40% among developers at firms with more than 5,000 employees.

Cursor also remains highly visible, with 69% awareness, but its work adoption now sits at 18%, tied with Claude Code. The story here is not that Cursor is fading. It is that the market is getting more selective, and awareness alone is no longer enough.

• Claude Code: 57% awareness, 18% work adoption, 91 CSAT, 54 NPS
• GitHub Copilot: 76% awareness, 29% work adoption, 40% adoption in companies with 5,000+ employees
• Cursor: 69% awareness, 18% work adoption
• Claude Code adoption in the US and Canada reached 24%

This is the clearest signal in the report: developers are not merely adopting AI. They are sorting tools by output quality, fit, and trust. The old assumption that one vendor can own the whole workflow is weaker now.

Chatbots still matter, but agents are moving in

Chat interfaces are still a big part of the picture. JetBrains found that 28% of developers used the ChatGPT chatbot for coding and development tasks at work. That is higher than the adoption of several specialized tools and shows that plain-language prompting remains a practical entry point for many teams.

Other chatbot usage is smaller but still meaningful. 8% of developers used Gemini, and 7% used Claude’s chatbot. Meanwhile, OpenAI’s coding agent Codex had 27% awareness and 3% work adoption in the January 2026 data, though JetBrains notes that this was collected before the public launch of the Codex desktop app and the promo inside ChatGPT.

Google Antigravity is the surprise entry. Launched in November, it reached 6% adoption by January 2026. That is a fast start for a new editor, especially in a market where developers already have habits, subscriptions, and keyboard shortcuts baked in.

JetBrains’ own tools also have a measurable foothold. JetBrains AI Assistant and Junie are used by 11% of developers combined, with AI Assistant at 9% and Junie at 5%.

• ChatGPT chatbot: 28% work usage
• Gemini chatbot: 8% work usage
• Claude chatbot: 7% work usage
• OpenAI Codex: 27% awareness, 3% work adoption
• Google Antigravity: 6% adoption by January 2026

The pattern is easy to read: chatbots remain the simplest on-ramp, but agents and editor-integrated tools are where serious usage is moving. Developers want help where the code already lives, not in a separate tab they have to keep switching back to.

JetBrains is betting on open agent workflows

JetBrains is not just reporting the data. It is also shaping its product strategy around it. The company says it wants an open ecosystem where developers can choose different agents for different tasks, rather than being locked into one vendor’s stack.

That idea shows up in several products. JetBrains IDEs can use Claude Agent and OpenAI Codex in the AI chat, and they can access many other agents through the Agent Client Protocol. JetBrains says developers can also use Codex with an OpenAI API key or a ChatGPT subscription.

JetBrains Central is the company’s control plane for agent-driven software production. It adds governance, cloud runtimes, and a shared semantic layer so agents can understand code organization at a system level. That is a big claim, but the product direction is clear: manage agents like infrastructure, not like toys.

Air, currently in public preview, goes further. It is a dedicated environment for running multiple agents in parallel inside isolated Docker containers or Git worktrees. JetBrains says this lets agents understand symbols, commits, and methods without touching the main working copy.

Junie CLI, now in beta, brings that model to the terminal. It is LLM-agnostic, supports bring-your-own-key setups, and lets developers switch between OpenAI, Anthropic, Google, and Grok models. That flexibility is not a small feature. It is the sort of thing teams ask for once they start comparing cost, latency, and output quality across vendors.

• JetBrains AI Assistant: 9% regular usage
• Junie: 5% regular usage
• JetBrains AI Assistant plus Junie: 11% combined usage
• Air supports isolated Docker containers and Git worktrees
• Junie CLI supports OpenAI, Anthropic, Google, and Grok via BYOK

My read: JetBrains is betting that the next competitive layer is orchestration. If developers start using multiple agents for different jobs, the winner will be the platform that makes them easy to compare, control, and audit.

What this means for teams buying AI tools

The survey is a reminder that AI adoption at work is already normal, but tool choice is still in motion. Copilot remains huge in enterprise environments. Claude Code is growing fast and getting strong reviews. ChatGPT still has wide usage because it is easy to reach for. New editors like Antigravity can still break through if they offer enough speed and convenience.

For engineering leaders, the practical question is no longer whether to allow AI coding tools. It is which ones fit the team’s workflow, security rules, and budget. A browser chatbot may be fine for ad hoc help, while an agent inside the editor may matter more for repetitive code changes and repo-wide tasks.

The next data point to watch is organizational adoption, not just individual usage. JetBrains says it will continue tracking agent workflows and related adoption challenges in its upcoming Developer Ecosystem Survey 2026. If the January numbers are any guide, the real competition now is between tools that help developers think and tools that help them ship.

So the question for teams is simple: are you standardizing on the tool everyone knows, or the one your developers will still choose after the novelty wears off?