Rulecatch SDK

Official open-source packages for Rulecatch — AI coding analytics, monitoring, and rule enforcement for Claude Code.

Track every AI coding session in real-time: tool calls, file edits, token usage, costs, and rule violations. Privacy-first with zero-knowledge encryption.

Packages

| Package | Description | npm | |---------|-------------|-----| | @rulecatch/ai-pooler | CLI + hooks for tracking AI coding activity | | | @rulecatch/mcp-server | MCP server — query violations and rules from Claude Code | |

Quick Start

Monitor AI activity (no account needed)

npx @rulecatch/ai-pooler monitor --no-api-key

See what your AI coding assistant is doing in real-time — tool calls, file edits, tokens, costs. Zero setup.

Full setup (with dashboard)

# 1. Install and configure
npx @rulecatch/ai-pooler init

# 2. Start the live monitor
npx @rulecatch/ai-pooler monitor

# 3. (Optional) Add MCP server for in-IDE rule violations
# Add to ~/.claude/mcp.json:
# { "mcpServers": { "rulecatch": { "command": "npx", "args": ["-y", "@rulecatch/mcp-server"] } } }

What It Tracks

• Tool calls — every Read, Write, Edit, Bash, Grep, Glob with success/failure and I/O size
• Token usage — input/output tokens per model, running totals
• Cost — real-time cost estimation by model (Opus, Sonnet, Haiku)
• Code changes — lines added/removed via git diff
• Session timeline — start, end, duration, active time
• Rule violations — custom rules enforced across your team

Privacy

All personal data (file paths, emails, project names) is encrypted on your machine before it leaves. We use AES-256-GCM with PBKDF2 key derivation (100k iterations). Your encryption key never leaves your device.

Read more about our privacy architecture

How It Works

Claude Code hooks → local buffer → flush to API → dashboard
                                                 ↓
                              rule engine → violation alerts

1. Hooks capture events from Claude Code (zero token overhead — hooks run in bash)
2. Buffer stores events locally as JSON files
3. Flush sends batched events to the Rulecatch API (encrypted)
4. Dashboard shows analytics, trends, and team insights
5. Rules are checked in real-time and violations surface in the MCP server

Links

• Website
• Dashboard
• Documentation

License

MIT