What is agent-canvas-ai-model?

Agent Canvas Model is a browser-based planning tool for organizing the design of an AI agent before building it. It borrows the visual style of a business model canvas, but the blocks are tailored to AI agent topics like governance, agent behavior, and MCP integration rather than general business strategy. According to the README, the canvas is split into 10 planning blocks arranged across 3 phases: governance, agent design, and the MCP layer. The whole thing is delivered as HTML, which means it can be opened locally from a folder, hosted on a static site like GitHub Pages, or used in an offline setup without needing a server or backend of its own. The tool includes autosave through the browser's localStorage, so filled-in canvas data persists as you work without manual saving. If you want to share a specific state of the canvas with someone else, it supports URL hash sharing, which encodes the canvas contents into a link. For moving data between sessions or devices, it supports JSON import and export, and it also offers a Word export option for people who prefer working with documents. Getting started involves cloning or downloading the repository and opening the main HTML file in a browser, or hosting the folder on a static web host such as GitHub Pages. From there, you fill in the 10 planning blocks across the 3 phases, let autosave track your progress, and export or share when the plan is ready. The README also notes the framework aims for accessibility aligned with WCAG 2.2 AA, and mentions both cloud and on-prem editions exist for different deployment preferences, though it does not go into detail on how those editions differ. The project is released under the GNU GPL v3.0 license.