Tools for people who write prompts and ship text
Twelve free utilities for working with large language models and raw text β token counting, structured prompt building, API cost estimation, regex testing, format conversion and diffing. Everything runs locally in your browser, so nothing you paste is uploaded anywhere.
AI Token Counter
Estimate tokens and cost for GPT, Claude and Llama style tokenisers.
π§±Structured Prompt Builder
Assemble a prompt from role, task, context, format and constraints.
π΅LLM API Cost Calculator
Compare monthly spend across models from tokens and request volume.
πContext Window Calculator
How many tokens you have left, and what fits in the remainder.
πRegex Tester
Test patterns with live match highlighting and capture groups.
πCSV β JSON Converter
Convert both directions with type detection and custom delimiters.
πText Diff Checker
See exactly what changed between two versions, line by line.
πURL Slug Generator
Turn headlines into clean, SEO-friendly URL slugs.
πText Case Converter
camelCase, snake_case, kebab-case, Title Case and more.
πMarkdown β HTML
Convert Markdown to HTML and back without a build step.
βοΈLorem Ipsum Generator
Placeholder text in paragraphs, sentences or words, no Latin clichΓ©s only.
π§ΉDuplicate Line Remover
Deduplicate, sort and clean lists of lines in bulk.
Why these tools run entirely in your browser
Prompts are often not public. They contain product roadmaps, customer data, unpublished code and internal strategy. Pasting that into an online tool that sends it to a server means letting a third party read it, and in many cases store it. Every tool here processes text with JavaScript inside your own tab. There is no account, no upload endpoint and no analytics on content β open the network panel and you will see no request carrying your text.
About token counts
Every language model splits text differently. GPT-style tokenisers use byte-pair encoding, Claude uses its own scheme, and the same sentence can differ by ten percent or more between them. The token counter here uses a well-established heuristic β roughly one token per four characters of English, adjusting for whitespace and punctuation β which lands within about ten percent for typical prose but is less accurate for code, numbers and non-Latin scripts. For billing-grade precision, use your provider's official tokeniser; for planning and sanity checks, the estimate is enough.
What good prompting actually changes
Prompting advice is often presented as magic wording, but the measurable improvements come from structure rather than phrasing. Specifying the output format removes ambiguity about what you want back. Giving the model relevant context instead of asking it to guess reduces hallucination. Setting constraints β length, tone, what to exclude β prevents the failure modes you already know about. The structured prompt builder exists because these five elements are the ones that reliably matter, and assembling them deliberately beats rewriting the same paragraph five times.
Controlling API cost
Most cost surprises in production come from three places: input tokens growing because a retrieval step dumps too much context, output tokens nobody bounded with a max_tokens setting, and retry logic that silently duplicates requests. The cost calculator and context window calculator together let you model all three before you ship. Caching prompt prefixes and trimming retrieved context usually cut spend more than switching to a cheaper model does.
How to use this site
- Before you call an API β count tokens and check the context window, so you know what fits and what it costs.
- While writing a prompt β build it from parts, then paste the result into your model of choice.
- While handling data β convert CSV to JSON, test a regex, diff two versions, clean up a list.
- While publishing β generate a clean slug and convert Markdown to HTML.
All twelve tools are free, work on mobile, and require no signup. Bookmark the ones you use most.