2 followers
What do you think? …
Last quarter I spent $340 on Claude API for one project. When I looked at
the logs, ~50% of my tokens were going to clarification follow-ups —
prompts that weren't specific enough the first time.
So I built EcoToken: it rewrites your prompts to be denser, then tracks
the actual tokens + dollars you saved across runs. Works with Claude,
GPT-4o, Gemini, or any LLM (BYO key for $5/mo, free tier with 10 runs).
Live at: ecotoken.dev
Free tier needs no card.
Happy to answer any technical questions about the optimization approach
(it's a mix of whitespace compression + Claude Haiku rewriting + per-
project calibration).
Last quarter I spent $340 on Claude API for one project. When I looked at
the logs, ~50% of my tokens were going to clarification follow-ups —
prompts that weren't specific enough the first time.
So I built EcoToken: it rewrites your prompts to be denser, then tracks
the actual tokens + dollars you saved across runs. Works with Claude,
GPT-4o, Gemini, or any LLM (BYO key for $5/mo, free tier with 10 runs).
Live at: ecotoken.dev
Free tier needs no card.
Happy to answer any technical questions about the optimization approach
(it's a mix of whitespace compression + Claude Haiku rewriting + per-
project calibration).