regex-mcp

MCP server for testing, explaining, debugging, and generating regular expressions. Works with any MCP client — Claude Code, Cursor, VS Code Copilot, Windsurf, and more.

Never leave your editor for regex again.

Features

• Test — Run patterns against text with highlighted matches, capture groups, and coverage stats
• Explain — Token-by-token breakdown with nested group indentation
• Validate — Syntax checking with common pitfall warnings (greedy traps, anchor issues, unescaped dots)
• Replace — Find and replace with capture group references ($1, $2)
• Debug — Visual match markers with smart hints when patterns fail (case, anchors, whitespace, multiline)
• Generate — Natural language to regex for 25+ common patterns (email, URL, UUID, date, phone, etc.)

Install

Claude Code

claude mcp add @muhammadalishahzad/regex-mcp -- npx -y @muhammadalishahzad/regex-mcp

Claude Desktop

Add to ~/Library/Application Support/Claude/claude_desktop_config.json:

{
  "mcpServers": {
    "regex": {
      "command": "npx",
      "args": ["-y", "@muhammadalishahzad/regex-mcp"]
    }
  }
}

Cursor

Add to .cursor/mcp.json:

{
  "mcpServers": {
    "regex": {
      "command": "npx",
      "args": ["-y", "@muhammadalishahzad/regex-mcp"]
    }
  }
}

VS Code Copilot

Add to .vscode/mcp.json:

{
  "servers": {
    "regex": {
      "command": "npx",
      "args": ["-y", "@muhammadalishahzad/regex-mcp"]
    }
  }
}

Windsurf

Add to ~/.codeium/windsurf/mcp_config.json:

{
  "mcpServers": {
    "regex": {
      "command": "npx",
      "args": ["-y", "@muhammadalishahzad/regex-mcp"]
    }
  }
}

Tools

regex_test

Test a pattern against input text. Returns highlighted matches, positions, capture groups, and coverage stats.

Pattern: \d+
Input:   "Order #123 has 4 items at $99.50"

Found 4 matches (8 chars, 25.0% of input)

Highlighted: Order #[123] has [4] items at $[99].[50]

Match 1:
  Text:     "123"
  Position: 7–10
  Length:   3

Match 2:
  Text:     "4"
  Position: 15–16
  Length:   1

regex_explain

Token-by-token breakdown with group indentation for nested patterns.

Pattern: (?:https?://)([\w.-]+)(?:/(\w+))?

Breakdown:
  (?:          → start non-capturing group
    h          → literal 'h'
    t          → literal 't'
    t          → literal 't'
    p          → literal 'p'
    s          → literal 's'
    ?          → optional (zero or one)
    :          → literal ':'
    /          → literal '/'
    /          → literal '/'
  )            → end group
  (            → start capture group
    [\w.-]     → any character in [\w.-]
    +          → one or more times
  )            → end group
  (?:          → start non-capturing group
    /          → literal '/'
    (          → start capture group
      \w       → any word character (a-z, A-Z, 0-9, _)
      +        → one or more times
    )          → end group
  )            → end group
  ?            → optional (zero or one)

regex_validate

Check syntax and detect common pitfalls.

Pattern: ^.*foo.bar$
Flags:   g

Valid regex pattern.

Flags:
  g — global — find all matches

Warnings (3):
  1. Unescaped '.' matches ANY character. Did you mean '\.'?
  2. '.*' is greedy — it matches as much as possible. Consider '.*?'
  3. Using anchors (^/$) with global flag (g). Add 'm' for per-line matching.

regex_debug

Visual debugging with smart hints when patterns don't match.

Pattern: /^Hello/g
Input:   "hello world" (11 chars)

Result: No matches.

Hints:
  1. Case mismatch: With the 'i' flag, the pattern matches 1 time(s).

When matches are found, shows visual markers:

Found 2 matches:

  "Error on line 42: timeout. Error on line 99: crash."
   ^^^^^^^^^^^^^^^                ^^^^^^^^^^^^^^^^^^^^^

Match 1 [0–15]: "Error on line 42"
  Context: [Error on line 42]: timeout. Err...

Match 2 [27–52]: "Error on line 99: crash."
  Context: ...timeout. [Error on line 99: crash.]

regex_replace

Find and replace with capture group support.

Pattern:     (\w+), (\w+)
Input:       "Doe, John"
Replacement: $2 $1

Replacements: 1

Before:
Doe, John

After:
John Doe

regex_generate

Generate regex from natural language descriptions. Supports 25+ common patterns.

Description: "email address"

Best match:
  Pattern: [a-zA-Z0-9._%+-]+@[a-zA-Z0-9.-]+\.[a-zA-Z]{2,}
  Description: Matches most common email address formats

Available categories: email, URL, IPv4, phone, date (ISO/US), time, hex color, UUID, number, decimal, HTML tag, credit card, ZIP code, MAC address, slug, camelCase, snake_case, password validation, JSON key, Markdown link, file extension, semver, and more.

Development

git clone https://github.com/MuhammadAliShahzad/regex-mcp.git
cd regex-mcp
npm install
npm run build

Test with the MCP Inspector:

npm run inspector

Run the server directly:

node build/index.js

How it works

regex-mcp runs as a local stdio process — no network calls, no API keys, no LLM dependencies. All regex operations use the native JavaScript RegExp engine. Pattern explanation uses a built-in tokenizer, not AI.

License

MIT