Code & Text Tools — 9 Free Browser Utilities

Best free code & text tools for 2026

Code & text tools sit between writing and shipping — every diff before commit, every regex before prod, every markdown render before publish runs through one of these utilities. The 9 tools in this category cover the operations that show up dozens of times per day in real engineering workflows: regex pattern testing with live capture-group display, two-document diff with line-level granularity, AI-token counting for prompt budgeting (GPT-4o, Claude Sonnet, Gemini), and the escape/unescape variants needed when shuttling strings between JSON, HTML, JavaScript, and shell. All run client-side via the browser's native regex engine, native JSON.stringify, and tokenizer libraries shipped as static JS — no input is ever uploaded.

What's the difference between regex testing and regex explaining?

A regex tester takes a pattern and a sample string and shows whether the pattern matches, what each capture group caught, and the full vs first match. Use it when you have a pattern and want to confirm it does what you intended. A regex explainer takes a pattern alone and translates it into plain English ("matches one or more digits, then a hyphen, then 3-4 word characters"). Use it when reading someone else's regex or auditing a complex production pattern. Both tools live in this cluster — Regex Tester for the former, Regex Explainer for the latter.

How do I count GPT-4 or Claude tokens online without sending my prompt to an API?

The AI Token Counter ships the same tokenizer libraries (cl100k_base for GPT-3.5/4, o200k_base for GPT-4o, Anthropic's tokenizer for Claude, Gemini SentencePiece) compiled to WebAssembly. Tokenization runs locally — your prompt text never reaches OpenAI, Anthropic, or Google servers. Verify with DevTools → Network tab: zero outgoing requests during count. This matters because production prompts often contain customer data or unreleased features; sending them to OpenAI's token counter API is the same data exposure as sending them to the model itself.

Code & text cluster — pick the right tool

Which code & text tool should you use?

Code and text tools cover the day-to-day grind: testing a regex you'll never remember, comparing two configs, escaping strings for JSON or HTML, counting tokens before sending a prompt to GPT or Claude. The right choice depends on the format you're wrestling with and whether you need deep analysis (regex explainer) or a quick transform (case converter).

• Test and debug a regex pattern with live highlighting and capture groups (JS flavor) — use Regex Tester & Debugger.
• Translate a regex to plain English when inheriting unfamiliar patterns — use Regex Explainer.
• Compare two text blobs side by side (configs, log diffs, JSON responses) — use Text Diff Checker.
• Write/preview Markdown with GitHub-flavored tables and code blocks — use Markdown Editor & Preview.
• Convert pasted Word/Google Docs content to clean Markdown — use Word to Markdown Converter.
• Convert variable names between camelCase, snake_case, kebab-case, PascalCase — use String Case Converter.
• Escape strings for JSON, HTML, JavaScript, URL, CSV, or SQL — use String Escape & Unescape.
• Count characters, words, lines, bytes for tweet limits, SMS, OG titles — use Character Counter.
• Estimate AI API costs before sending a prompt to GPT/Claude — use AI Token Counter. Catches expensive prompts early.

Common code & text mistakes to avoid

• Greedy regex that matches too much. .* grabs everything to the end of line. Use .*? (lazy) or a more specific character class.
• Forgetting to escape the dot in regex. . matches any character — use \. when you mean a literal period (e.g. matching example\.com).
• Diffing two files that have different line endings. CRLF vs LF causes every line to show as changed. Normalize line endings first.
• Mixing escape contexts. JSON-escaping " as \" then re-encoding for HTML produces double-escaped junk. Encode once, in the final destination format.
• Trusting the 4-chars-per-token heuristic for non-English. CJK, code, and emojis can be 1 char per token or worse. Always count tokens in the actual tokenizer for cost-sensitive prompts.
• Counting characters for Twitter/X without weighting URLs. URLs count as 23 chars regardless of length. Use a counter that knows the rules.