Verified project record

basketikun/infinite-canvas

infinite-canvas is an open-source visual workbench that unifies AI image generation, reference editing, conversational assistants, and prompt libraries on a multi-canvas interface. It connects to user-configured OpenAI-compatible API endpoints from the browser for text-to-image, image-to-image, audio, and video generation.

Project overview

It combines canvas orchestration, AI media generation, reference editing, conversation, and prompt libraries into a single visual interface, and supports a local Canvas Agent that connects to Codex or Claude Code via MCP so an agent can operate the current canvas.

Project type: MCP · Workflow · Image & Vision
Use cases: Documents & Office · Image Processing · Automation
Deployment: Refer to project documentation
License: AGPL-3.0

Best for

Creators who want a unified visual workspace combining AI image generation, reference editing, conversational assistants, and prompt libraries in a single interface.
Developers who need an extensible canvas with a TypeScript SDK for custom plugins and agent operation via MCP.

Key capabilities

Supports multi-canvas projects, node drag-and-drop scaling, connections, minimap, undo/redo, and import/export.
Directly connects to OpenAI-compatible interfaces from the browser frontend for text-to-image, image-to-image, reference image editing, text Q&A, audio, and video generation.
Converses and generates images around selected and upstream nodes, inserting results back into the canvas.
Connects to Codex or Claude Code via a local Canvas Agent, allowing the agent to operate the current canvas via MCP.
Provides a Codex app plugin that automatically registers MCP and attempts to launch the local Agent upon installation.
Supports dynamic installation, enabling, updating, and uninstallation of remote node plugins via URL, including a TypeScript SDK for custom development.
Allows customization of image/video API call methods to flexibly adapt to various relay stations and self-built services.
Connects directly to multiple open-source GitHub projects from the browser frontend and caches them in IndexedDB.

Limitations and risks

The project is currently in the development stage and does not guarantee backward compatibility for historical data.
Various local storage formats may be directly adjusted without warning.

Getting started

Clone the repository, run docker compose up -d, access http://localhost:3000, and configure your OpenAI-compatible Base URL and API Key.

Evidence and sources

GitHub project description
README
README
README
README

Reviewed at: 2026-07-30T03:55:21Z

Open original GitHub project