TrimPrompt — Save 80% on LLM Tokens

Your AI coding agent sends package-lock.json (500KB), binary files, and your .env secrets to the LLM API — at full token price.

TrimPrompt intercepts CLI output and compresses it before the LLM sees it. Zero config.

What it does

What your agent reads	Before	After	Saved
`package-lock.json`	36,097 tokens	125 tokens	99.7%
`dashboard.png`	250,542 tokens	113 tokens	99.9%
`ls` (4,542 files)	75,109 tokens	436 tokens	99.4%
`find` (951 files)	44,495 tokens	1,486 tokens	96.7%
`.env` with AWS keys	Sent as-is	[REDACTED]	Blocked

C — Compression: 27 command-specific filters for git, cat, ls, find, docker, kubectl, and more
S — Security: Auto-redacts AWS keys, API tokens, JWTs, database strings, private keys (20 patterns)
A — Audit: Per-command tracking, AI tool auto-detection, savings dashboard

No config files. No API keys. No code changes. Everything runs locally.

Claude Code, Cursor, GitHub Copilot CLI, Gemini CLI, Aider, Windsurf, Cline, Continue.dev

git, cat, ls, tree, find, grep, npm, docker, kubectl, cargo, pytest, pip, go, and more (22 commands)