Bring Your Own AI: Free MCP, Skills & Local Image Models

October opened the v2.5 series and changed how AI works in zStudio. The MCP server became free on every plan, including Community: connect an AI tool you already use, such as Claude (desktop or Claude Code), Antigravity, Cursor, VS Code, Windsurf or Gemini CLI, and it can build scenes in the editor on your own subscription without using zStudio credits. Help → Connect AI (MCP) shows the server address and a ready-to-paste setup snippet for each tool. AI Skill Libraries let you teach the AI your studio's own conventions, such as how you structure popups or name templates. Add a folder of Markdown files, a single skill file, or a starter library to edit; the AI loads only the skills it needs for each task. If you already have scenes, the dialog includes a prompt that has an AI tool study them and write the skill files for you. Local Image Models brought image generation to your own computer, free and with no account, credits or internet connection. Stable Diffusion 1.x, 2.x and XL, FLUX, Z-Image and Qwen-Image models load from .safetensors or .gguf files, multi-part models in the ComfyUI folder layout are found automatically, and text-to-image, image-to-image and in-place reskin all work. AI tools connected over MCP can use your local models too. The MCP toolset itself grew: an AI can now duplicate a container as a new template, place new instances of existing templates and save the scene, and less common operations are reachable through a single run_action tool that the AI can look up with describe_actions. If the default MCP port is already taken, zStudio picks the next free one. Renaming a template now reliably keeps its children's animations, even when a child's name contains the template's name. The website gained a dedicated AI page explaining the whole setup.