HN Showcase
Back to browse
I built a proxy that cuts LLM costs 40-60% – no AI involved

I built a proxy that cuts LLM costs 40-60% – no AI involved

by christalingx·Mar 3, 2026·2 points·1 comment
Visit ProjectView on HN

AI Analysis

●●SolidSolve My ProblemSlick

Prompt compression API cuts token bills 40-60%, integrates in two lines.

Strengths
  • •Monkey-patch mode requires zero code changes, immediate savings for existing apps
  • •42% average token reduction across 2.4M+ real API calls with measurable beta traction
  • •5ms latency overhead and infrastructure-agnostic design work with any LLM provider
Weaknesses
  • •Compression quality depends entirely on proprietary algorithm with no transparency into technique
  • •Competes with prompt caching (OpenAI/Claude), context windows, and structured outputs—all free alternatives
Target Audience

Backend developers, AI/ML engineers, SaaS founders using OpenAI/Claude APIs

Similar To

OpenAI prompt caching · Claude token optimization · LangChain prompt compression

Similar Projects

Developer Tools●●Solid

NadirClaw, LLM router that cuts costs by routing prompts right

If you're burning through Claude/OpenAI credits, this is a low-friction stopgap: it classifies prompts in ~10ms and routes trivial tasks to cheaper/local models while reserving premium APIs for complex work. The agentic-task detection, reasoning-aware routing, session pinning and context-window fallback are practical touches that avoid mid-thread model bouncing and 429 failures. It isn't reinventing the space (OpenRouter and others exist), but it's focused on real-world cost tradeoffs and drop-in compatibility.

Solve My ProblemNiche Gem
amirdor
113mo ago