Home

snip

The extensible LLM token optimizer. CLI proxy that filters shell output before it reaches your AI coding assistant's context window.

AI coding agents burn tokens on verbose shell output that adds zero signal. snip sits between your AI tool and the shell, filtering output through declarative YAML pipelines before it reaches the context window.

132 filters, 20 pipeline actions, 15 CLI commands, 13 AI integrations.

Measured on a real Claude Code session: 128 commands, 2.3M tokens saved, 99.8% average savings.

Pages

• Installation - curl install, Homebrew, binary, from source
• Integration - Claude Code, Cursor, Codex, Windsurf, Cline, OpenCode, OpenClaw, standalone
• Gain Dashboard - Token savings reports and analytics
• Learn - Detect CLI error-correction patterns, generate agent rules
• Filters - 132 built-in filters, pipeline actions, custom filters
• Filter DSL Reference - Complete YAML schema and 20 action reference
• Configuration - TOML config, environment variables
• Architecture - Design decisions, internals, data flow
• Contributing - Development setup, adding filters, conventions

Quick Start

# Quick install (macOS/Linux)
curl -fsSL https://raw.githubusercontent.com/edouard-claude/snip/master/install.sh | sh

# Or via Homebrew
brew install edouard-claude/tap/snip

# Or with Go
go install github.com/edouard-claude/snip/cmd/snip@latest

Then hook into your AI tool:

snip init                        # Claude Code (default)
snip init --agent cursor         # Cursor
snip init --agent codex          # OpenAI Codex
snip init --agent pi             # Pi
snip init --agent windsurf       # Windsurf
snip init --agent cline          # Cline / Roo Code
snip init --agent copilot        # GitHub Copilot
snip init --agent gemini         # Gemini CLI
snip init --agent kilocode       # Kilo Code
snip init --agent antigravity    # Antigravity
snip init --agent grok           # Grok

Commands

Command	Description
snip <command>	Run command through filter pipeline (implicit)
snip run	Run command through snip filter pipeline (use -- to separate)
snip check	Check if a command would be filtered (use -- to separate)
snip init	Install agent integration (default: claude-code)
snip hook	Handle agent PreToolUse/shell hook
snip hook-audit	Show recent hook activity (set SNIP_HOOK_AUDIT=1 to log)
snip gain	Show token savings report
snip cc-economics	Show financial impact of token savings by API tier
snip discover	Scan sessions for missed filter opportunities
snip learn	Detect CLI error-correction patterns in sessions
snip verify	Run inline filter tests (--require-all to enforce coverage)
snip config	Show current configuration
snip trust	Trust project-local filter file(s) by SHA-256 hash
snip untrust	Remove filter file(s) from the trust store
snip proxy	Passthrough without filtering (optional -- separator)
snip inspect	Code quality checks for snip's own source

How It Works

┌─────────────┐     ┌─────────────────┐     ┌──────────────┐     ┌────────────┐
│ Claude Code │────>│ snip intercept  │────>│ run command  │────>│   filter   │
│  runs git   │     │  match filter   │     │  capture I/O │     │  pipeline  │
└─────────────┘     └─────────────────┘     └──────────────┘     └─────┬──────┘
                                                                       │
                    ┌─────────────────┐     ┌──────────────┐           │
                    │   Claude Code   │<────│ track savings│<──────────┘
                    │  sees filtered  │     │  in SQLite   │
                    └─────────────────┘     └──────────────┘