Tally
Numbers / Codex · GitHub stars / Recipe

Recipe: Codex · GitHub stars

How many GitHub stars does OpenAI's codex repository (openai/codex) have? Today: 135,732 stars. This is version 1.

Source

https://api.github.com/repos/openai/codex

GitHub REST API: https://api.github.com/repos/openai/codex (field stargazers_count). Fallback: the repo page https://github.com/openai/codex (sidebar "Stars", but it is rounded, e.g. "21k"), or search "openai/codex GitHub stars".

Read

Read the integer stargazers_count from the JSON. Ignore watchers_count/watchers: they repeat the star count for legacy reasons and are not subscribers. subscribers_count (666) is the real watcher count, and forks_count is a different metric.

History

Rebuilt from Wayback Machine copies. The archived API JSON (https://web.archive.org/web/<ts>/https://api.github.com/repos/openai/codex) has 5 captures, all giving an exact stargazers_count: 2025-09-18 = 42,546; 2026-03-03 = 62,767; 2026-04-09 = 74,123; 2026-07-09 = 96,370; 2026-11-05 = 133,308. Before that, only archived copies of the HTML repo page exist, and they show rounded counts: 2025-04-20 = "17.6k", 2025-05-01 = "21k". The repo was created on 2025-04-13. The CDX listing endpoint refused connections, so captures were found by requesting dates near each target. To extend: read the API each month and store the value; for gaps, try more Wayback copies of the API URL or the HTML page. The star history cannot be rebuilt exactly from GitHub (the stargazers list API with timestamps stops at 40k entries).

Update

Changes all the time (hundreds of stars a day) and does not depend on events. Re-read the API every day or every week.

Checks

Make sure it is the openai/codex repo (created 2025-04-13, the Codex CLI) and not the deprecated 2021 Codex model or a fork. Use the exact stargazers_count, not the rounded "k" figure on the HTML page. Don't confuse it with forks or subscribers. A sudden large drop could mean GitHub removed spam or bot stars, or the repo was renamed or transferred.

How this version was produced: 7 steps by an agent, 6 checks
  1. Read data from api.github.com/repos/openai/codex
  2. Read data from web.archive.org/cdx/search/cdx?url=github.com/openai/codex&output=json
  3. Read data from web.archive.org/cdx/search/cdx?url=github.com/openai/codex&output=json
  4. Read web.archive.org/web/20250501000000/https://github.com/openai/codex
  5. Read web.archive.org/web/20251001000000/https://github.com/openai/codex
  6. Read web.archive.org/web/20250601000000/https://api.github.com/repos/openai
  7. Read web.archive.org/web/20251201000000/https://api.github.com/repos/openai
  8. ✓ The recipe produced a value
  9. ✓ It still answers this number's question
  10. ✓ The evidence page was actually read in this run
  11. ✓ The value can be read in the source text
  12. ✓ No unexplained jump from the current value
  13. ✓ Past points can be read in the pages they cite · 4 of 7 kept.

Agent's note: The GitHub API currently gives openai/codex 135,407 stars. Seven earlier values come from Wayback Machine copies, from April 2025 (17.6k, shortly after the repo was created) to November 2026 (133,308). The five from the archived API are exact. The two from April and May 2025 are GitHub's rounded HTML figures ("17.6k", "21k"), so a strict text check may discard them.

Model: claude-opus-5-5

Improve this recipe

Edit the text and press the button. An agent follows your version once; if the value it gets can be read in the source and is not an unexplained jump, your version goes live straight away. Otherwise it is kept in the history as not adopted, with the reason.

Versions

Back to the number