The race to integrate AI coding assistants into developer workflows just got a new contender with the launch of CleanCode, a canvas-based Agentic Development Environment (ADE) designed to facilitate human-AI collaboration. Developed by chen-985211 and showcased on Hacker News, the tool brings together coding agents like Claude Code and Codex, terminal tasks, and long-running services into a unified visual workspace. The project aims to shift the developer's role from executing commands to guiding intent, effectively offloading the complexity of tool operation to the AI agents themselves.

System 1 Meets System 2

CleanCode’s architecture is explicitly modeled after the dual-process theory from Daniel Kahneman's 'Thinking, Fast and Slow.' The platform assigns 'System 1' tasks—fast, intuitive responses and goal-setting—to the human user, while delegating 'System 2' work—focused analysis, planning, and reasoning—to the AI agents. This division of labor allows developers to express what they want and observe the results on the canvas, retaining control over key judgments while the agents handle the intricate steps of learning software interfaces, breaking down tasks, and troubleshooting scripts.

Multi-Agent Orchestration and MCP Integration

At its core, CleanCode supports 33 different Coding Agent CLIs, including industry heavyweights like Codex and Claude Code, with automatic detection of installed tools on the user's machine. The platform introduces 'Native Agent Collaboration' via the CleanCode MCP (Model Context Protocol), enabling agents to operate the canvas directly, build workflows, and even create other agents. This interoperability is supported by branch workspaces that leverage Git worktrees, allowing developers to run multiple changes in parallel, each with its own isolated canvas, terminals, and agent sessions.

Preview Status and Technical Requirements

Currently, CleanCode is available as a Preview for macOS, Windows, and Linux, though it lacks official developer certificate signing. Users on macOS must manually bypass Gatekeeper restrictions and verify SHA256 checksums from the GitHub Releases page to launch the application. For those interested in contributing or running from source, the project requires Node.js version 24 or higher and pnpm 10, with development workflows supported by standard checks like typechecking and testing via pnpm scripts.

Key Takeaways

  • CleanCode unifies 33 coding agents, including Claude Code and Codex, into a single visual canvas environment.
  • The platform uses a 'System 1/System 2' model to separate human intuition from AI analytical execution.
  • Native MCP integration allows agents to manipulate the canvas and create sub-agents autonomously.
  • Branch workspaces use Git worktrees to enable parallel development streams with isolated agent contexts.
  • The current release is an unsigned Preview, requiring manual verification and Gatekeeper bypass on macOS.

The Bottom Line

CleanCode offers a compelling vision for the future of AI-assisted development by treating agents as collaborative partners rather than just autocomplete tools. If the preview stability holds, this canvas-based approach could redefine how developers orchestrate complex, multi-agent workflows.