CanvasCode

CanvasCode is a native macOS application that provides an infinite canvas where multiple AI agents can work simultaneously. Each agent operates in its own panel with a real terminal, allowing all processes to remain visible at once. Users can open as many agents as needed, each running different models such as Codex, Claude, or Grok.

CanvasCode is designed for users who need to manage multiple AI agents simultaneously, such as developers, designers, or teams working on complex projects that require coordination between different AI models and tools.

CanvasCode includes a voice assistant called Nina that allows users to manage agents, generate images, or query project status without using the keyboard. Users can speak commands to open agents, split tasks, install skills, or organize the canvas.

Yes, CanvasCode supports MCP tools and skills from a public registry, which can be installed and enabled per project. Agents can also install these tools directly from within the app, allowing for seamless integration with external systems.

Agents in CanvasCode coordinate themselves by splitting workstreams, communicating with each other, and opening additional agents when needed. The canvas organizes itself, and a built-in memory system tracks each agent’s progress and contributions.

Each agent panel in CanvasCode includes a built-in browser that displays the results of their work, such as screenshots and console output. This eliminates the need to switch applications and keeps all outputs visible within the canvas.