caveman vs codex
The two are close in size: 113.8k stars for caveman, 130.1k for codex. Over the days we have tracked them caveman moved +7.6% and codex +29.1%, so codex is growing faster right now.
caveman leads on context depth and noise control and customization and model freedom. codex does not take any axis by a clear margin.
Stars and commit dates come from our own daily tracking. The six axes are read off each project's documentation by our review pipeline, so they describe what a project says about itself, not what we measured in its code.
Where they stand today
Compresses both what the agent writes (a caveman-style skill) and what it reads (a local proxy) to cut token usage while keeping originals accessible.
- Stars
- 113.8k
- Tracked growth
- +7.6%
- Maturity
- ●●●●●
- Last commit
- 1d ago
- Language
- Go
- License
- Apache-2.0
- Cost to run
- Free to install — uses your AI provider's billing
A lightweight, local-first terminal coding agent that you can install and run with one command while signing in to ChatGPT or using your OpenAI API key.
- Stars
- 130.1k
- Tracked growth
- +29.1%
- Maturity
- ●●●●●
- Last commit
- 25m ago
- Language
- Rust
- License
- Apache-2.0
- Cost to run
- Sign in with a ChatGPT plan or use your OpenAI API key
Six axes, head to head
Each axis runs 0 to 5. The label under a score is what that project's own docs claim, not a category average.
| Axis | caveman | codex |
|---|---|---|
Context depth How much of your codebase it sees before it answers: the open diff, the diff plus related files, or the whole repository. | ●●●●● Diff + related files | ●●●●● Diff/file only |
Noise control How it keeps output volume down — severity thresholds, deduplication, incremental runs over new commits only. | ●●●●● Basic filters & dry-run | ●●●●● None mentioned |
Customization How far it bends to your team: custom rules, prompts, style guides, per-path config. | ●●●●● Configurable rules & modes | ●●●●● No customization |
Privacy Whether your code stays on your own infrastructure: fully local, self-hostable, or cloud API only. | ●●●●● Self-hostable | ●●●●● Cloud (ChatGPT/API) |
Model freedom Whether you can point it at any provider, or it is wired to one. | ●●●●● Many specific providers | ●●●●● OpenAI only |
Setup ease What it takes to get a first useful run out of it. | ●●●●● One-command install | ●●●●● One-command install |
Which one to pick
Pick caveman if…
Token-savings — use Caveman when you want an easy, local install that meaningfully reduces model input and output tokens without changing correctness.
- Context depth: Diff + related files (3/5 against 1/5)
- Noise control: Basic filters & dry-run (3/5 against 1/5)
- Customization: Configurable rules & modes (4/5 against 1/5)
- Model freedom: Many specific providers (3/5 against 1/5)
Pick codex if…
Easy setup — one-command install to run a local CLI coding agent that integrates with your ChatGPT plan or OpenAI API key.
What people want from each one
JuliusBrussee/caveman
openai/codex
Hacker News: Replit's new Code LLM: Open Source, 77% smaller than Codex, trained in 1 week drew 891 points and 220 comments.
Questions people ask
Is caveman better than codex?
caveman leads on context depth and noise control and customization and model freedom. codex does not take any axis by a clear margin. caveman is worth picking when token-savings — use Caveman when you want an easy, local install that meaningfully reduces model input and output tokens without changing correctness.
Which of caveman and codex keeps my code private?
caveman: Self-hostable (4/5). codex: Cloud (ChatGPT/API) (3/5).
What does each one cost to run?
caveman: Free to install — uses your AI provider's billing. codex: Sign in with a ChatGPT plan or use your OpenAI API key.
Full profiles: JuliusBrussee/caveman and openai/codex. Everything else in CLI coding agents.