Status: [LIVE]

Endpoints found:
https://billyzhonguom.github.io/llm

Source: https://github.com/BillyZhongUOM/llm-cost-calculator

# TokenTally — LLM API cost calculator

Live: **https://billyzhonguom.github.io/llm-cost-calculator/**

A free, client-side calculator that estimates what LLM API usage actually costs, and compares providers on your exact workload. It models the three things most "tokens times price" calculators skip:

1. **Prompt caching** — a repeated prefix billed at the cache-read rate (~0.1x input), after the provider's write premium or first-call cost.
2. **Batch discounts** — the ~50% async discount, where a provider offers one.
3. **The agent / chatbot context tax** — every turn re-sends the whole history, so cost compounds turn by turn. TokenTally sums each turn explicitly and shows how much caching claws back. A flat per-step estimate can be off by 2x or more on a long agent run.

No backend, no sign-up, nothing leaves your browser except the share link you choose to copy.

## Why it exists

The market is full of token-times-price comparison tables, and they all stop at the easy part. The cost that surprises teams is the agent loop, where a 12-step run does not bill 12 small prompts, it bills a prompt that grows toward 20,000 tokens, again and again. No calculator surveyed modelled that turn by turn. This one does, and it turns the result into advice ("your toggles save $X/mo", "caching is costing you here, turn it off") rather than a bare number.

## Scenarios

| Scenario | Models | Volume unit |
|----------|--------|-------------|
| Single call | one prompt in, one answer out | calls / month |
| Chatbot | multi-turn thread, context grows each turn | conversations / month |
| Agent | tool loop, context piles up each step | runs / month |
| RAG | template + retrieved chunks + question | queries / month |
| Batch | template + per-item, async discount | items / month |

## Architecture

Vanilla HTML, CSS and ES modules. No framework, no build step.

```
index.html     structure, SEO, JSON-LD
styles.css     design system (amber-on-slate instrument, light + dark)
engin