AI & LLM Tools

Client-side tools for prompt work and structured LLM output. Count tokens across models, estimate context-window fit, lint Mustache/Jinja/LangChain templates, diff or minify prompts, validate model JSON against a schema, generate function-calling schemas, and compare embedding vectors—all in your browser with no provider API calls.

LLM Token Counter

Count GPT tokens across models. 100% client-side.

Context Window Calculator

See how much of a model context window a prompt plus expected output would use.

Prompt Diff Checker

Compare two prompt versions with a line or word diff. 100% client-side.

LLM JSON Schema Validator

Validate structured LLM JSON output against a JSON Schema. Client-side Ajv.

Function Calling Schema Generator

Infer an OpenAI or Anthropic tool schema from sample JSON.

Prompt Template Linter

Lint mustache, Jinja, and LangChain prompt templates for delimiter and variable issues.

Embedding Cosine Similarity Calculator

Compute cosine similarity, dot product, and Euclidean distance for two embedding vectors.

Prompt Minifier

Strip markdown and extra whitespace from prompts to save tokens.

Frequently Asked Questions

Do these tools call OpenAI or Anthropic?

No. Token counts, diffs, and schema checks run entirely in JavaScript. Nothing is sent to an LLM provider.

Are token counts exact for every model?

OpenAI GPT-family counts use a real BPE tokenizer and are exact. Claude, Gemini, Llama, and DeepSeek results are labeled Estimated.

Where do I validate structured model output?

Use the LLM JSON Schema Validator with a local Ajv build. Pair it with the Function Calling Schema Generator when you need a draft tool schema from sample JSON.