DevImprint.

Tools → Compare

Codex vs Claude Code vs Cursor: what 363K developers’ public commits show

Among 363,366 people with detectable AI coding-tool attribution in public repositories, 340,547 have Claude Code attribution, 46,204 have Cursor attribution, and 6,966 have Codex attribution.

90% of people with Codex attribution also have Claude Code attribution.

52% of people with Cursor attribution also have Claude Code attribution.

These counts reflect which tools leave detectable public attribution, not market share or total use.

Source snapshot 8 November 2026, complete ranked population of 363,366 people. Not a judgement of which tool is better.

Codex vs Claude Code vs Cursor usage, side by side

Developers in the complete ranked population, by detected tool
ToolDevelopersShareActive, 30 daysOnly this toolMedian commits
Claude Code340,54793.7%89,283 (26%)310,80812
Cursor46,20412.7%12,375 (27%)21,89611
Codex6,9661.9%3,723 (53%)65541

Share is of all people with any detected tool; shares overlap because one person can have several tools. Median commits is the median 30-day count of detected AI-assisted commits among active people, from every tool they use, not only this one.

Do developers use more than one?

Read each row as: of the developers with the row's tool, the share who also have the column's tool. The overlap is asymmetric, because the populations differ in size.

Overlap between tools among people in the complete ranked population
Developers with……also Claude Code…also Cursor…also Codex
Claude Code—7% (24,250)2% (6,264)
Cursor52% (24,250)—3% (1,250)
Codex90% (6,264)18% (1,250)—

Codex vs Claude Code vs Cursor: how the tools differ

What each tool is and where it runs
CodexClaude CodeCursor
Made byOpenAIAnthropicAnysphere
InterfaceCommand line, IDE extension, and a cloud agentTerminal first, with IDE extensions, desktop and webThe editor itself: completion, chat and an agent
RunsOn your machine, or as cloud tasks in their own environmentOn your machine, or in cloud sessionsAs a desktop editor, with background agents available
PricingSee OpenAISee AnthropicSee Anysphere

Choosing between Codex, Claude Code and Cursor

The deciding question is usually where you want to work. Codex suits the terminal, or handing tasks to a cloud agent that works in parallel. Claude Code suits working from the terminal and delegating multi-step changes across a repository. Cursor suits staying inside an editor with AI completion, chat and an agent beside the code.

Many developers do not choose. 90% of developers with Codex attribution also have Claude Code attribution, so pairing the tools is common among the most active public users.

Why these numbers are not market share

DevImprint counts tools from public attribution in repository history. Tools differ in whether, and how often, they leave that attribution, and developers can switch it off. A tool that attributes its work by default will appear far more often than one that usually leaves no trace, whatever their real user bases are.

The data therefore answers “how many developers show detectable use of each tool in public code”, not “how many developers use each tool”. Private repositories, inline completion and unattributed use are invisible to it. See what this leaderboard cannot see and the methodology. GitHub Copilot, Windsurf and similar tools rarely leave detectable attribution, so they are not compared here.

Codex vs Claude Code vs Cursor: common questions

Do more developers use Codex, Claude Code or Cursor?

Claude Code. 340,547 developers show Claude Code attribution in public commits, against 46,204 for Cursor and 6,966 for Codex. Tools differ in how often they leave public attribution, so this describes public, attributed use rather than overall popularity.

Do developers use Codex and Claude Code together?

Yes, often: 90% of developers with Codex attribution also have Claude Code attribution (6,264 people). Commit data cannot tell whether they use both at once or moved from one to the other.

What is the difference between Codex, Claude Code and Cursor?

Codex is a coding agent from OpenAI; Claude Code is an agentic coding tool from Anthropic; Cursor is an AI code editor from Anysphere, built on VS Code. Codex suits the terminal, or handing tasks to a cloud agent that works in parallel; Claude Code suits working from the terminal and delegating multi-step changes across a repository; Cursor suits staying inside an editor with AI completion, chat and an agent beside the code.

Is Codex, Claude Code or Cursor better?

Public commit data cannot say which is better: commit counts do not measure quality, productivity or fit for a task. Choose by where you want to work, and note how many developers use more than one.

Codex vs Claude CodeCursor vs Claude CodeCodex vs Cursor

Cite this statistic

DevImprint recorded 6,264 people with detected attribution for both Codex and Claude Code for Codex and Claude Code. Population: people in the complete ranked population. Source snapshot: 2026-11-08T20:28:18.634744+00:00 (d5ada2cdab024e900790b6073b24a564c20dccc8e8748dea35d39a81c87e6499). 33.2% of known repositories scanned; provisional. Methodology: 2026-09-14. Page: https://devimprint.com/compare/codex-vs-claude-code-vs-cursor/. Current data: https://devimprint.com/leaderboard.json

Data & coverage

Figures come from the complete Parquet export for snapshotd5ada2cdab02 and update with each publication.