PlateSpinner – A Kanban board that orchestrates AI coding agents

Name: PlateSpinner – A Kanban board that orchestrates AI coding agents
Availability: InStock
Author: mordymoop

by mordymoop·Mar 6, 2026·1 point·0 comments

Visit Project View on HN

AI Analysis

●●SolidShip ItSolve My Problem

Kanban UI for headless Claude Code—three-phase pipeline with cost tracking and autoclicker mode.

Strengths

•Three-phase pipeline (propose → plan → execute) with read-only sandboxing for early stages reduces risk before full write access.
•Autoclicker autonomous mode with AI judge deciding what to do next, budget caps, and configurable parallelism—genuinely thoughtful safety design.
•Built-in cost tracking and diff viewer provides tangible visibility into token spend and code changes per task.

Weaknesses

•Relies on spawning external CLI tools (Claude Code, Codex, Gemini) rather than direct API calls—adds complexity and dependency friction.
•Competes directly with Cursor, Continue, GitHub Copilot, and Anthropic's Claude editor—all offer AI-assisted coding without local agent orchestration overhead; unclear what real projects gain from kanban overhead vs. IDE integration.

Post Description

PlateSpinner is a local web app that manages AI coding agents through a kanban board. You point it at a project directory, describe what you want, and it spawns headless Claude Code / Codex / Gemini sessions to generate tasks, plan implementations, and write code.

The workflow is: Propose → Plan → Execute. Generation and planning run in read-only mode. Execution gets full write access, creates branches, commits changes, and captures diffs for review. There's also an "autoclicker" mode where an AI judge continuously decides what to do next, propose, plan, or execute, so you can just watch.

Built with React + Express + WebSocket. It doesn't call any cloud APIs directly — it shells out to the CLI tools the same way you would from a terminal.