Research
Claude Code vs Cursor vs Copilot, as employers write them
Every comparison of these tools, Cursor vs Claude Code included, measures what developers use. This counts what employers put in the requirements — 8,239 postings at 90 AI companies’ own job boards, read on 26 August 2026.
- boards name Claude Code
- 46
- name Cursor
- 16
- name Copilot
- 9
The order is not the one the discourse would predict
Claude Code is named at 46 of 90 boards, Cursor at 16 and GitHub Copilot at 9. Copilot almost certainly has the largest install base of the three and comes last here by a distance.
That inversion is the finding, and it is not a measure of quality or of how many people use each one. It is a measure of what an employer considers load-bearing enough to write into an advert, and those are different things.
An editor is a personal choice. A company that is happy for you to use Cursor has no reason to say so in a job description, and one that has bought Copilot licences for everyone has even less — ubiquity is exactly what stops a tool being mentioned. What lands in the requirements is the thing a role is expected to build with or against: an agent a team has wrapped workflows around, or a protocol its product exposes.
The same pattern runs through the full tool ranking, where MCP outranks every editor on the list.
Every assistant, side by side
Ranked by how many distinct companies name it, not by total mentions. A company’s own board is excluded from its own tool’s count, so a maker cannot top its own table by describing itself.
| Assistant | Companies | Postings |
|---|---|---|
| Claude Code | 46 of 90 | 174 postings |
| Cursor | 16 of 90 | 31 postings |
| GitHub Copilot | 9 of 90 | 16 postings |
| Codex | 8 of 90 | 62 postings |
| Devin | 1 of 90 | 89 postings |
| Windsurf | 1 of 90 | 11 postings |
Why the table counts companies and not mentions
Devin appears in 89 postings but at only 1 company. One employer writing it into most of its adverts produces a large mention count and tells you almost nothing about the industry.
Sorted by raw mentions it would sit near the top of this page. A tool named once at forty companies is a far stronger signal than one named forty times at a single company, so the column that decides the order is the number of companies.
What this does not tell you
These are 90 AI companies’ own boards, so this describes hiring inside that industry rather than software hiring generally. It is also the population where these tools are most likely to appear at all — so read the shares as an upper bound, not an average.
And a mention is a mention. This counts a tool named anywhere in a posting, whether in the stack, the responsibilities or a nice-to-have. It cannot distinguish “you will use this every day” from “familiarity a plus”, and it does not try to.
Common questions
- Cursor vs Claude Code: which do employers ask for?
- Claude Code, by a wide margin, in job postings. It is named at 46 of 90 AI companies' job boards against Cursor's 16. That is a measure of what employers write down, not of what developers install — and the two diverge for a specific reason: an editor is a personal choice, so a company has no cause to require one, while an agent a team builds workflows around ends up in the requirements.
- Claude Code vs Copilot: which do employers ask for?
- Claude Code, by a wide margin: 46 of 90 boards name it, against 9 for GitHub Copilot, which appears in 16 of 8,239 postings. It is worth reading that carefully: Copilot has by far the largest install base of anything here, and it still barely appears. Ubiquity is exactly what stops a tool being written into a job advert.
- Which AI coding tools appear most in job descriptions?
- Among coding assistants: Claude Code at 46 of 90 boards, Cursor at 16 of 90 boards, GitHub Copilot at 9 of 90 boards. Across all tool categories the leaders are cloud infrastructure and MCP rather than any assistant.
- Does naming a tool mean the job requires it?
- No. This counts postings that name a tool anywhere in their text — in the stack, the responsibilities, or as a nice-to-have. It measures what an employer thought was worth writing down, which is a weaker claim than a requirement and a more useful one than asking developers what they have installed.
- How was this measured?
- Every open role on 90 AI companies' public applicant tracking systems was fetched on 26 August 2026 and matched against a fixed vocabulary. A company's own board is excluded from its own tool's count, so Cognition's postings do not inflate Devin. Ranking is by distinct companies rather than total mentions.