Agent Q Canvas: Coordinating Local AI Coding Agents on a Spatial Canvas

Code repo: https://github.com/Shellishack/agent-q-canvas

Agent Q Canvas started from a real workflow problem: when I work across several products on the same day, local coding agents can become scattered across terminals, repositories, project folders, browser tabs, and context windows.

When a personal computer has too many windows and tabs open, it becomes easy to lose context and lose track of what is actually being worked on. For my own workflow, I can already have five or six Codex tabs open, plus VS Code, browser previews, terminals, documentation, and project folders.

The more useful AI coding agents become, the more coordination work appears around them. Someone still needs to route tasks, review blocked states, approve actions, watch previews, and understand which project is moving forward.

Agent Q Canvas is an Electron app for coordinating local coding agents on a spatial canvas. The current prototype supports multiple canvases, dashed project groups, draggable project frames, agent nodes, local task queues, a global command box, a project jump list, and a preview drawer.

The direction is to make local agent work feel visible and organized instead of fragmented. Instead of spreading work across disconnected windows, the product puts agent terminals, browser previews, task routing, project context, and approvals onto one visual canvas.

That canvas can support one person managing multiple projects or a team coordinating work together. The goal is a more organized and efficient operating surface for AI-assisted development.

The near-term roadmap includes real local agent terminals, browser preview nodes, adapters for Codex, OpenCode, Claude Code, and Gemini, persistent project and task state, git worktree isolation, and a command center for approvals and blocked agents.

This is still an early prototype, but the product direction is clear: AI coding workflows need better operating surfaces. Agent Q Canvas is my attempt to build one for the way I already work.