# SaveToken

> API that shrinks logs, JSON, and tool/MCP schemas before they reach an LLM provider. Up to 50% fewer input tokens on structured payloads (based on in-house testing); a `verdict` field reports whether the reduction passed a programmatic proof (round-trip equality or a reference-model check) or left the text unchanged. Never calls the LLM itself — returns the reduced text for you to send to Claude, OpenAI, or any other provider with your own key.

## What it does

- Uniform JSON arrays: re-encoded so a repeated key set is sent once, not once per row. Checked with round-trip equality before shipping.
- Tool/MCP schemas: minified losslessly by default; verbose descriptions trimmed only with logged evidence that agent tool-selection still works afterward.
- Logs & tracebacks: library-internal frames dropped; app frames and the exact root-cause line survive byte-for-byte.

## What it does not do

- Does not summarize or paraphrase — never generates new text.
- Does not call any LLM provider on your behalf.
- Does not help with human prose — no benefit claimed there.
- Does not store request or response content — only token counts, dollars saved, verdict, and mode per call.

## API

- `POST /v1/optimize` — body `{"text": "..."}` or `{"messages": [...]}`, header `Authorization: Bearer <key>`. Returns `{"optimized": "...", "verdict": "WIN"|"NEUTRAL"|"PASS_THROUGH", "tokens_in": n, "tokens_out": n, "dollars_saved": n}`.
- `POST /v1/keys/free` — mints a free key, 100 calls/day, no card required.
- `GET /v1/usage`, `GET /v1/savings` — usage history and lifetime dollars saved for your key.
- `POST /v1/support` — file an issue, no key needed. Body `{"message": "...", "request_id": "...", "contact": "..."}` (only `message` required).

## Pricing

Free tier: 100 calls/day. Paid: $49/mo, higher quota, billed via Paddle. Cancel anytime.

## Who it's for

AI agencies running client workflows, n8n/Make.com automators, Claude Code/Cursor power users, teams shipping AI agents, anyone piping logs or JSON into an LLM.

## Links

- API reference: /docs
- Help center (getting started, billing, usage limits, error troubleshooting; machine-readable): /help.md
- Pricing (machine-readable): /pricing.md
- Terms: /terms.html
- Privacy: /privacy.html