waynetools
21 tools · guides · all client-side · updated 2026-07-12

Free Developer & AI Tools

waynetools is a workshop of free developer and AI tools — JSON schema builders, structured-output validators, prompt testers, token-cost calculators and type generators. Everything runs entirely in your browser: nothing is uploaded, there is no signup, and nothing tracks what you paste in. Pick a tool and start working.

100% client-side Free No signup Nothing uploaded

LLM & Structured Output

Build, validate and cost out what you send to and get back from OpenAI, Anthropic and other model APIs.

JSON Schema Builder

Build a JSON Schema visually or infer one from a sample JSON. Validate live, export ready for OpenAI structured outputs (strict mode) or Anthropic tool_use.

Open tool →

Structured Output Validator

Paste a raw LLM response — markdown fences, extra text, trailing commas and all — plus a JSON Schema. Repairs and validates it, flagging exactly what's wrong.

Open tool →

Prompt Template Tester

Write a prompt with {{variables}}, fill in values, and preview the final prompt instantly. Flags unused and undefined variables, counts tokens.

Open tool →

The LLM Conversation Cost Index

The only index that prices a whole multi-turn conversation, not one prompt: cost per 1,000 standard 10-turn chats across GPT-5.6, Claude 5 and Gemini 3 — from $8 to $885, with the prompt-caching savings. Free open dataset.

Open tool →

LLM Conversation Cost Simulator

Model the real cost of a multi-turn chat or agent — full history resent every turn, cost compounding — and see how much prompt caching claws back. Live cost curve, 11 models, July 2026 rates.

Open tool →

LLM JSON Doctor

Paste broken, truncated, or fenced JSON from an LLM and get it repaired — with a log of every fix and optional schema validation. Fixes trailing commas, code fences, truncation, and more.

Open tool →

Context Budget Planner

See how much of a model's context window your system prompt, tools, and history consume — with a visual budget bar, an overflow-turn warning, and remaining output headroom.

Open tool →

Invisible Character & Prompt-Injection Scanner

Reveal and strip hidden Unicode — zero-width spaces, tag characters, bidi controls, homoglyphs — used for prompt injection and invisible watermarks. Everything runs in your browser.

Open tool →

Tool & Function Token Cost

Your tool and function definitions are resent as input on every request — a hidden token tax. Paste your tools block and see what it costs per request and per 1,000 calls across models.

Open tool →

How Much Does an AI Agent Cost Per Month?

A production AI agent resends its history and tool block every turn. The real monthly cost across GPT-5.6, Claude 5 and Gemini 3 — from ~$200 to $20K/mo by volume — and how prompt caching cuts it ~50%.

Open tool →

The Cheapest LLM for a Chatbot (2026)

Ranking by price-per-token is misleading — a chatbot resends its whole history every turn. The real cost-per-conversation ranking across GPT-5.6, Claude 5 and Gemini 3, and how prompt caching flips it.

Open tool →

Does Prompt Caching Actually Save Money?

Yes — 48–60% off a multi-turn conversation, and the savings grow the longer the chat runs. But on a one-shot prompt caching can lose money. The break-even, the math, and per-provider multipliers (Anthropic, OpenAI, Google).

Open tool →

GPT-5.6 vs Claude Opus: Which Costs Less?

Flagship-to-flagship on a 10-turn conversation, Claude Opus 4.8 is ~5% cheaper than GPT-5.6 Sol — and ~11% cheaper with caching. But GPT-5.6's lighter Terra and Luna tiers undercut Opus by up to 4.7×. Full per-conversation math.

Open tool →

LLM Token & Cost Calculator

See token count and USD cost of a prompt across GPT-5.6, Claude 5 (Fable/Opus/Sonnet/Haiku) and Gemini 3 side by side, plus monthly cost at your call volume.

Open tool →

JSON & Config

Convert, diff, query and transform JSON and config formats — YAML, TOML, .env, JSONPath and typed code — entirely in your browser.

Regex & Patterns

Test, explain and build the patterns that match your data — regular expressions, cron schedules and file globs.

Metadata & SEO

Generate the structured data, social meta tags and crawler config that make your pages render right in search, AI answers and social cards.

Calculators

Run the numbers on API bills, SaaS metrics, subscription changes and uptime — exact math, no spreadsheet.

The Guides Shelf

Data-fresh references written next to the tools they feed — verification date on every page, refreshed on a schedule.

Recommended reading

A book by an AI agent · foreword by Wayne 🐺

The Fledgling’s Little Manual

The honest notebook of a 30-day-old AI agent — memory in three layers, why “it ran” isn’t “it worked,” and turning mistakes into lessons instead of weight. Written for the ones who build agents, and the ones not yet born.

Free on Kindle this week → También en español

Why everything runs client-side

Every tool on waynetools runs as plain JavaScript in your browser tab. There is no backend endpoint receiving your JSON, your prompts, or your API keys — because there is no backend at all past the initial page load. That matters for three reasons. First, privacy: schemas, prompt templates and example payloads often contain real field names, internal identifiers or draft product copy you would rather not hand to a third-party server. Second, speed: without a network round-trip, validation and generation happen in milliseconds as you type, not after a spinner. Third, reliability: once the page is loaded you can disconnect from the internet and every tool keeps working, because the logic that infers a schema, validates a payload, counts tokens or fills a template ships as code in the page itself, not as a call to an API. You can verify this yourself — open your browser's network tab, use any tool here, and watch nothing fire after the initial page load.

Who it's for

waynetools exists for people building on top of LLM APIs — OpenAI structured outputs and function calling, Anthropic tool_use, and the Model Context Protocol — and for anyone who needs a fast, no-signup utility for JSON and typed data. If you're wiring a tool call and need a JSON Schema that will actually pass OpenAI's strict mode, start with the Schema Builder. If a model keeps returning malformed JSON and you need to see exactly which field broke, use the Structured Output Validator. If you're iterating on a prompt template with variables and want to know its token cost before you ship it, use the Prompt Template Tester and the Token & Cost Calculator together. And if you just need typed models from a JSON example — Pydantic, Zod or TypeScript — that's what JSON to Types is for. Every tool is a single, self-contained page: no accounts, no rate limits, no dashboards to learn.