Codex leads measured use, OpenCode leads GitHub attention — the shares invert

Asked:

List all popular open-source agents. What is their user share and Github stars share?

No census covers every popular open-source agent, and project-reported installs and downloads are not comparable. This report therefore scopes the share comparison to autonomous open-source AI coding agents with an explicit worldwide work-adoption percentage in the JetBrains Developer Ecosystem Survey 2026 (May–July 2026, 15,509 professional developers) — a complete measurable set of two: Codex (16% adoption) and OpenCode (7%). That scoping does not mean only two popular open-source agents exist; notable others are named in the appendix.

Codex OpenCode — the crossing lines mark the inversion between professional use and community stars

Within the two-agent comparable set, Codex holds 69.6% of normalized user share on 16% worldwide work adoption — blog.jetbrains.com

OpenCode's 7% adoption gives it 30.4% of user share, yet its roughly 200,000 GitHub stars are 63.6% of the pair's combined 314,500 — x.com

Adoption rates are non-exclusive — 90% of the surveyed developers used coding agents at work weekly and most use several tools — heise.de

The 16% Codex figure covers the broader Codex product family, while the ~114,500 stars belong only to the open-source openai/codex CLI repository — byteiota.com

The two-agent comparable set

Agent	Adoption %	User share %	Stars ≈	Star share %	Repository	Caveat	Source
Codex	16%	69.6%	114,500	36.4%	openai/codex	JetBrains measures the broader Codex product family; stars are for the open-source openai/codex CLI repo	blog.jetbrains.com
OpenCode	7%	30.4%	200,000	63.6%	opencode	Star count from a dated late-August 2026 observation; approximate	x.com

Excluded from the share math

Cline, Aider, OpenHands, Roo Code, Continue, Goose, Gemini CLI and Qwen Code are all notable open-source coding agents.

They are excluded from the share calculations because the same representative survey published no adoption percentage for them. Their popularity is reported in incompatible units — installs, downloads, monthly actives, VS Code installs — and mixing those with survey adoption would manufacture a misleading market share.

Method: shares computed only within the two autonomous open-source coding agents with an explicit worldwide work-adoption percentage in the JetBrains Developer Ecosystem Survey 2026 (fielded May–July 2026; 15,509 professional developers per Heise). Adoption rates are non-exclusive, so user shares are normalized within the pair (16+7=23 points). GitHub star counts are approximate late-August 2026 snapshots and date-sensitive; combined stars 314,500. Other open-source agents were cut for lack of comparable survey data.