Describe it. Get the regex.

What this tool understands

The translator recognizes three kinds of phrases. Combine them freely.

Whole-pattern intents (use a known good pattern)

These map directly to vetted patterns from our library:

• email address · email
• URL · web address
• IP address · IPv4
• UUID · GUID
• hex color · color code
• phone number · US phone
• ZIP code · postal code
• date · ISO date · YYYY-MM-DD
• time · HH:MM
• semver · version number
• strong password
• MAC address
• domain name · hostname

Quantified atoms (how many of what)

• any digit → \d
• any letter → [a-zA-Z]
• exactly 4 digits → \d{4}
• at least 8 letters → [a-zA-Z]{8,}
• between 2 and 5 word characters → \w{2,5}
• 3 uppercase letters → [A-Z]{3}

Anchors (where in the string)

• starts with hello → ^hello
• ends with .com → \.com$
• starts with foo, then 3 digits → ^foo\d{3}

Why "rule-based, not AI"?

AI regex tools (autoregex.xyz, several "AI regex generators") produce patterns that look plausible but are often subtly wrong. The cost of a wrong regex in production is high — data loss, validation holes, security issues.

This translator is intentionally narrower. It only emits regex it knows is correct, by mapping recognized phrases to vetted patterns. When it doesn't understand, it says so — instead of confidently guessing.

For anything outside this tool's grammar, use the explainer to verify a regex written by hand, or paste an AI-generated pattern into the explainer to check what it actually does.