Claude Pilot – Claude Code is powerful. Pilot makes it reliable
TDD enforcement and context preservation for Claude Code workflows, but hooks-on-every-edit pattern is established.
MCP server + embedded terminal that lets Claude Code see and edit your ComfyUI workflows
This repo actually hands an LLM live control over a ComfyUI graph — list node types, create/connect/delete nodes, tweak params, run queues and even view preview images — via an MCP server and an embedded terminal. The idea of treating the workflow as a JSON DAG so each edit maps to a tool call is smart and pragmatic; it's the kind of niche automation that'll save hours for people iterating image pipelines. My main caveat: giving a model direct edit rights raises obvious safety/permission questions and ties the UX tightly to Claude.
ComfyUI users, generative AI artists, prompt engineers, hobbyist ML practitioners and power users who build image-generation pipelines
TDD enforcement and context preservation for Claude Code workflows, but hooks-on-every-edit pattern is established.
Clever use of Claude Channels to push messages—nobody else bridged these two agents yet.
Multi-model debate is clever; execution wraps existing tools without core differentiation.
Loop driver + 15 slash commands for Claude Code, but orchestration over integration.
Hash-verified edits cut Claude editing tokens by 44%—measurable, specific, real win.
MCP server scans sites and generates Claude-ready fix prompts for direct editing.