Browse by type
21% less than Codex per solved task on Terminal-Bench 4. Same model on both sides.
21% cheaperthan Codex Terminal-Bench 4, per solved task, GPT-6.1 Sol: $1.21 vs $1.52, 90 of 192 runs solved both |
Levelwith Claude Code within noise Terminal-Bench 4, Opus 5.5: 97 of 180 runs solved vs 83 of 177; $5.19 vs $6.04 per solved task, 95% interval 0.70x to 1.02x |
Measured build, benchmark settings on, not the shipped app: the app ships the same harness with three benchmark-only settings turned off. Same model on both sides: GPT-6.1 Sol against Codex, Opus 5.5 against Claude Code. Our own runs, November 2026, at list prices. How we measured
Every coding agent in one window. Free on the subscriptions and keys you already have.
Download Caveman Code for Mac · Read the launch post

Caveman make your AI agent say less and read less. Code stay exact. Brain still big.
33.2% fewerinput tokens, whole sessions 6 synthetic large-tool-output tasks, 18 pairs, Aug 2026 engine, 18 of 18 answers right |
0.46×the cost of Opus at xhigh, with Auto 85 coding tasks, interim, score gap ±7 points, unresolved |
397 → 76tokens for one test run Blocks vs| tail -20 |
1.4 to 2.4× cheaperwith caveman-style output Adobe Research, eight models |
Cited by Adobe Research · A/B tested by JetBrains · Remade for Elasticsearch on Elasticsearch Labs
#1 on Hacker News · #1 on GitHub Trending · "No way this actually works." ThePrimeagen
Install · Six rocks · Proof · Say less · Read less · Auto · Blocks · Commands · Caveman Cloud
| Normal agent · 63 tokens | |
|---|---|
| > The reason your React component is re-rendering is likely because you're creating a new object reference on each render cycle. When you pass an inline object as a prop, React's shallow comparison sees it as a different object every time, which triggers a re-render. I'd recommend using useMemo to memoize the object. | > New object ref each render, so React re-renders. Wrap the prop in `useMemo`. |
Same fix. 63 token become 20. Brain still big.
Pick your club:
| Skill | Same answer | Tokens |
|---|---|---|
/caveman |
New object ref each render, so React re-renders. Wrap the prop in useMemo. |
20 |
/ultracave |
Inline object prop, new ref, re-render. useMemo. |
14 |
/megacave |
新參照致重繪。useMemo。 |
13 |
Token counts: tiktoken o200k.
npx @caveman-ai/cli@latest
One command. Mac, Linux, Windows. Caveman find your agents and your logins, show what it change. Press Enter. Done. Needs Node.js 22.13+.
After that, plain claude stays caveman'd. So do Codex, Gemini CLI, OpenCode, and 30+ more. It runs on your machine, with your keys and your Claude Pro/Max login. One rock. That it.
Had caveman before 4.0? Say @latest, or the old copy on your machine wins. Already set up? caveman setup opens the screen again. Only want shorter answers? Get just the skill.
New in 4.0: caveman is six parts. All on after install. Each has its own switch.
| Rock | What it does | Measured |
|---|---|---|
| Say less | Agent answers short. Code stays exact | 3% shorter than "Answer concisely." (ultracave 35%), n=10 |
| Read less | Logs, JSON, CSV, test output shrink before the model reads them | 33.2% fewer input tokens per session (6 synthetic large-tool-output tasks, 18 pairs, Aug 2026 engine) |
| Auto · new | Picks the model and the effort for each ask | 0.46× the cost of Opus at xhigh in an interim run (85 tasks); quality gap unresolved |
| Blocks · new | Code mode: agent keeps the scripts it writes and runs them again | 397 → 76 tokens for one test run |
| Waste fixes | Finds where your tokens go and fixes it. New: keeps the prompt cache warm while you pause | Not measured yet |
| Browse | Web pages come back compressed | 14.9% smaller full page than Playwright; 121 vs 15,704 tokens with a query, on one 200-row table |
caveman status shows what is on. caveman off browse turns one off. Only Auto needs an account, and it is free. 4.0.0 release notes
Outside labs first. Ours come with each rock below. Nothing rounded up. Red rows stay red.
| Who | Setup | Result |
|---|---|---|
| Adobe Research, CAVEWOMAN paper (cites this repo) | Caveman-style output, eight models, five datasets | Cost cut 1.4 to 2.4× per model, up to 3× |
| Elastic, Elasticsearch Labs | Caveman mode remade for Elasticsearch, eight live MCP scenarios | 63.6% fewer response tokens. "Zero information loss." |
| JetBrains | 86 real coding tasks, paired A/B | No measurable quality loss (p = 0.82). 8.5% fewer output tokens |
Two findings shaped caveman. Adobe: cavemanning your prompt makes answers worse, so caveman never touches your prompt. JetBrains: agent sessions are mostly code and tool calls, which the voice leaves alone. So caveman grew rocks that shrink what the agent reads and what it runs on.
A voice, not broken grammar. Every reply follows the same structure.
| Rule | What it means |
|---|---|
| Answer first | [thing] [action] [reason]. [next step]. No greeting, no "let me", no recap |
| One idea per sentence | Built on ASD-STE100, the controlled English of aircraft maintenance manuals: 20 words max, active voice, one term per thing |
| Meaning never dropped | Articles can go. not, never, no, only never go. Numbers and units stay exact |
| Payload verbatim | Code, commands, paths, and error messages untouched, character for character |
| No theater | No "me think". No chatter between tool calls. If caveman phrasing isn't shorter, plain wins |
| Knows when to stop | Security warnings, irreversible actions, and confused users get full sentences. Then grunt resumes |
| Your prompts stay yours | Never rewritten. Research say that backfire |
Example run, n=10, not a claim. Ten dev questions on claude-opus-5-5:
| Instruction | Output tokens |
|---|---|
| None | 6,983 |
Answer concisely. |
4,334 |
/caveman |
4,119 |
/ultracave |
2,693 |
"Be concise" is the real baseline; new models already know it. On top of it, /caveman cuts 3% more at the median and /ultracave 35% more. Measured on the 3.1.0 rules, not yet re-run. The rules cost about 1,160 tokens of input, and if you pay per request (GitHub Copilot), shorter answers save nothing. Harness · Every case where caveman loses
Only want this rock? The skill alone, no proxy:
npx skills add JuliusBrussee/caveman -g
Type /caveman ($caveman in Codex) if it doesn't start on its own. Say stop caveman to go back.
Other ways in: Claude app and phone, Claude Code plugin, Gemini, every agent at once, Windows, uninstall
Claude on the web, the phone app, or Cowork? No terminal. Add caveman to your Claude account once, in Customize > Plugins: steps. Not tested end to end yet.
# Claude Code plugin, auto-starts every interactive session, subagents too
claude plugin marketplace add JuliusBrussee/caveman && claude plugin install caveman@caveman
# Gemini CLI
gemini extensions install https://github.com/JuliusBrussee/caveman
# Every agent on your machine at once, plus the Claude Code statusline badge (Node.js 22.13+)
curl -fsSL https://raw.githubusercontent.com/JuliusBrussee/caveman/v4.2.0/install.sh | bash
Already inside Claude Code? Same rock, slash form. Type these where you talk to Claude:
/plugin marketplace add JuliusBrussee/caveman
/plugin install caveman@caveman
Windows, PowerShell 5.1+:
irm https://raw.githubusercontent.com/JuliusBrussee/caveman/v4.2.0/install.ps1 | iex
On npm 12 or newer, new npm block git install. One-liners above handle it themselves from v3.2.0 on. Old one-liner pinned to v3.1.0 or older still fail there: use one above, or this:
npx --allow-git=root -y github:JuliusBrussee/caveman#v4.2.0
Changed your mind: npx -y github:JuliusBrussee/caveman -- --uninstall (on npm 12+, add --allow-git=root too)
Only grabbed the skill? npx skills remove JuliusBrussee/caveman -g take it back out.
Install broke? Open your agent in this repo and say "Read INSTALL.md, install caveman for me." Agent fix own brain.
The voice shrinks what the agent says. This shrinks what it reads. Same answers: 18 of 18 right. 6 synthetic large-tool-output tasks, 18 pairs, Aug 2026 engine.
| File type | The file, through caveman | File saved | Whole Claude Code session, 3 runs | Session saved |
|---|---|---|---|---|
| CSV | 28,041 → 314 | 98.9% | 165,823 → 74,484 | 55.1% |
| Logs | 22,810 → 348 | 98.5% | 148,807 → 74,068 | 50.2% |
| YAML | 20,447 → 178 | 99.1% | 132,124 → 71,027 | 46.2% |
| Test output | 18,806 → 203 | 98.9% | 150,377 → 108,514 | 27.8% |
| JSON | 18,837 → 281 | 98.5% | 147,975 → 108,939 | 26.4% |
| HTML | 21,670 → 21,670 | none yet | 140,687 → 154,641 | 9.9% worse |
| All six | 130,611 → 22,994 | 82.4% | 885,793 → 591,673 | 33.2% |
The file is each test file through today's compressor alone. The session is the whole agent run, which also carries the system prompt and tools, so it moves less. Honest part: the session run is from August 2026 on an older engine, HTML has no compressor yet so it lost, and each test file hides one record in 61 to 72 KB of noise. Yours will vary. Headroom on the same suite: 15 of 18 right, 6.7% fewer tokens (interval crosses zero; harness not public). Method
Every original stays on your disk. The agent can pull it back any time.
New model in your agent's picker. Pick Auto and caveman chooses the model and the effort for each ask. Hard ask, big brain. Easy ask, save rocks.
| Setup | Tasks solved | Cost per try | Median time |
|---|---|---|---|
| Opus 5.5, xhigh effort, every ask | 44.3% | $7.83 | 35.2 min |
| Auto, Claude models only | 43.7% | $3.59 | 15.9 min |
| Auto, plus OpenAI and open models | 42.9% | $3.47 | 16.7 min |
Auto costs 0.46× what Opus 5.5 at xhigh costs. That is 54% saved, in under half the time. 85 real coding tasks from SWE-Interact and SWE-Together, one try each. Honest part: the run is half done, the 0.5-point score gap is too small to tell apart at this size (±7), and most of the saving is Auto lowering effort on easy asks. Already on Opus at medium? Expect less; not measured yet. Full table and limits
```bash caveman login # free account. Auto shows up in the model picker caveman providers ad
browse all types & interfaces →
$ claude mcp add caveman \
-- python -m otcore.mcp_server <graph>