Lexus2016

Cheap LLM Router — AI skill for Claude Code

AI community

Save Claude/Codex subscription tokens by routing context-only file reads through a cheaper LLM (DeepSeek, Kimi, Gemini Flash).

How to install Cheap LLM Router

This entry records only its repository, not the path inside it, so there is no exact command to give. Open Lexus2016/cheap-llm-router and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What Cheap LLM Router does

Save Claude/Codex subscription tokens by routing context-only file reads through a cheaper LLM (DeepSeek, Kimi, Gemini Flash). One-line install, works with any OpenAI-compatible model.

Alternatives in AI

System Prompts Leaks — Extracted system prompts from ChatGPT (GPT-5.4, GPT-5.3, Codex), Claude (Opus 4.6, Sonnet 4.6, Claude Code), G 38.6k ★
Optimization — ai-research-skills AWQ, GPTQ, GGUF, Flash Attention, bitsandbytes 5.4k ★
EchoBird — One-click install + model switch:Claude Code,Codex CLI (OpenAI), Grok Build (xAI), DeepSeek Harness, Kimi Code 3.1k ★

README

cheap-llm-router

**Save your Claude (or Codex) subscription tokens by letting a cheaper model do the file-reading for you.**

🌐 **Languages:** **English** · [Українська](README.uk.md) · [Русский](README.ru.md)

🤖 **Don't want to install this manually?** Paste this prompt to your AI agent (Claude Code, Codex CLI, Cursor, Gemini, Aider — anything that can read URLs and run shell commands):

*"Read and follow the* ***For AI agents (auto-install)*** *section in the README to install, configure, verify, and confirm `cheap` is running on this machine. Ask me only for my OpenRouter API key when you actually need it."*

The agent will pipx-install, prompt you for the API key once, install the rule into your `CLAUDE.md` / `AGENTS.md`, run a smoke test, and report back. Tested with Claude Code, OpenAI Codex CLI, Gemini CLI, and Cursor.

What it does, in plain words

When Claude Code or OpenAI Codex CLI opens files for you, every line of those files gets fed into your expensive subscription model. Five-file reads alone burn around 8 000 tokens. A few of those per day and you bump into the weekly limit.

`cheap-llm-router` adds one small command: **`cheap`**. It sends those files to a much cheaper model (DeepSeek, Kimi, Gemini Flash, your pick) and gets back a short summary — typically 600 tokens. Your expensive model receives the **answer**, not the raw bytes.

Same idea for the chat history: when you're about to compact a long session or update docs from what you just did, `cheap extract` summarises the session log so the expensive model doesn't have to re-read it.

Why you might want this

Hit the weekly limit less often. Or never.
Same workflow. You keep typing in Claude or Codex; cheap plugs in via a one-line rule in CLAUDE.md / AGENTS.md.
Any cheap model. Works with anything OpenAI-compatible. Default is deepseek/deepseek-chat-v3-0324 (~$0.005 per typical 4-file call, 164K-