Herman Geldenhuys

recursive-ai

Terminal UI toolkit for Claude Code — explore data with interactive panels, inject runtime artifacts onto a canvas viewer.

claude-code-sdk

claude-hooks-sdk

AI-native SDLC with structured knowledge capture and lifecycle management

Stack-based TUI framework with two modes: 'explore' (interactive panel apps) and 'canvas' (runtime artifact injection). Includes skills, library source, and example apps.

Human-friendly session names for Claude Code. Assigns memorable names like 'brave-elephant' to sessions, persists across /compact and /clear. v2.0: centralized storage, transcript tracking, machine namespacing, sesh CLI with transcript/machines/migrate commands.

YAML-based hook framework with built-in handlers. Includes session naming, turn tracking, dangerous command guards. Routes ALL hook events through a unified framework with environment variable injection.

Skill: Guide for implementing Claude Code hooks - bash scripts and TypeScript patterns for all 9 hook events.

Skill: Expert guide for using claude-hooks-sdk with HookManager, Logger, transforms, and advanced patterns.

Automatic session cost tracking, token usage monitoring, and performance metrics. Track every dollar and token spent across all sessions.

Unified conversation logging across all sessions with color-coded output. Perfect for multi-agent development and debugging.

Real-time anomaly detection with Slack/webhook alerts. Detect error spikes, token anomalies, and response time issues automatically.

Record all hook events to JSONL files for debugging, testing, and replay. Perfect for creating test fixtures from real sessions.

Track response times, SLA compliance, and tool performance. Monitor p95/p99 metrics and get alerted when SLAs are breached.

Comprehensive file change tracking with audit trails. Track every Write, Edit, and MultiEdit operation with timestamps and session IDs.

Stream events to a live web dashboard via SSE. Monitor Claude Code sessions in real-time with a beautiful web interface.