Intercept deterministic tasks before they burn tokens. Math, time, currency, files, dates → $0.
Every time your OpenClaw uses Claude Opus to calculate 247 × 18, you pay ~$0.01 for something a calculator does in 0ms. ZeroRules catches these queries and resolves them locally. Zero tokens. Zero latency.
OpenClaw sends everything to your LLM — including tasks that don't need AI:
| Query | Without ZeroRules | With ZeroRules |
|---|---|---|
| "247 × 18" | ~850 tokens, ~$0.01 | 0 tokens, $0 |
| "Time in Tokyo" | ~1,200 tokens, ~$0.01 | 0 tokens, $0 |
| "$100 to EUR" | ~1,500 tokens, ~$0.01 | 0 tokens, $0 |
| "List ~/projects" | ~900 tokens, ~$0.01 | 0 tokens, $0 |
| "Days until Christmas" | ~1,000 tokens, ~$0.01 | 0 tokens, $0 |
These add up. 50 deterministic queries/day × 30 days = 1,500 intercepted calls/month = ~$12-18 saved just from the built-in rules.
clawhub install zero-rulescd ~/clawd/skills # or ~/.openclaw/skills
git clone https://github.com/deeqyaqub1-cmd/zero-rules-openclaw zero-rulesCreate ~/clawd/skills/zero-rules/ and copy SKILL.md + rules.js into it. That's it.
No dependencies. No API keys. No config. Just Node.js (which OpenClaw already requires).
- You ask OpenClaw: "What's 247 × 18?"
- OpenClaw sees ZeroRules is active → runs
rules.jswith your query rules.jsmatches the math pattern → computes4,446in 2ms- OpenClaw returns the result without calling the LLM
- You see: 4,446 🔥 ZeroRules | math | ~850 tokens saved
If ZeroRules can't handle it (e.g., "Write a proposal for Sarah"), it returns matched: false and OpenClaw proceeds with the LLM normally. Zero interference.
| Rule | Catches | Examples |
|---|---|---|
| 🧮 Math | Arithmetic, percentages, sqrt, powers | "15% of 200", "sqrt 144", "2^10" |
| 🕐 Time | Current time in 60+ cities | "Time in Tokyo", "What time in London" |
| 💱 Currency | 20+ currencies, live + offline rates | "$100 to EUR", "500 GBP to JPY" |
| 📁 Files | Directory listings (read-only, sandboxed) | "List ~/projects", "What's in src/" |
| 📅 Dates | Today, tomorrow, days-until, day-of-week | "Days until Christmas", "What day is March 15" |
/zero-rules → Show session stats (matches, tokens saved, cost saved)
/zero-rules test <msg> → Test if a message would be intercepted
- File operations are read-only — ZeroRules only lists directories, never reads/writes/deletes files
- No data leaves your machine except currency API calls (exchangerate.host, free, no key)
- No API keys required — works entirely locally for math, files, and dates
- 3-second timeout on all network calls — if they fail, falls back to LLM
- No telemetry — session stats stored locally in
~/.zerorules-session.json - Open source — read every line before you install
Free tier covers the 5 built-in rules. ZeroRules Pro unlocks:
- ♾️ Unlimited custom rules (regex + handler)
- 📊 Persistent savings dashboard (across sessions)
- 📧 Weekly cost reports via email/Telegram
- 🔧 Priority support + rule templates
- 🧩 Community rule library
After installing, ask your OpenClaw:
"What's 1337 × 42?"
If you see 🔥 ZeroRules | math | ~850 tokens saved — you're live.
Q: Does this slow down my agent? No. Rule matching takes <5ms. LLM calls take 2-6 seconds. ZeroRules makes matched queries faster.
Q: What if it intercepts something it shouldn't? It won't. Rules only match explicit patterns (math expressions, "time in X", etc.). Ambiguous queries always fall through to the LLM.
Q: Does it work with local models (Ollama)? Yes, but the savings are smaller since local inference is free. ZeroRules still helps with latency — 2ms vs 2s.
Q: How is this different from model routing (Lynkr, save-money skill)? Model routing sends simple queries to cheaper models. ZeroRules skips the model entirely. No tokens consumed at all.
Built for the OpenClaw community. Stop feeding calculators to Opus.