Z-Image Studio

You can run Z-Image Studio as an MCP server to expose image-generation capabilities to local agents or other tools. It supports stdio for local automations, as well as Streamable HTTP and SSE endpoints for remote clients. This makes it easy to integrate image generation, model listing, and history access into automated workflows while keeping content consistent across transports.

How to use

To use the MCP server, run the server alongside the web UI or separately so agents can request image generation, list available models, and fetch history. The server provides a stdio interface for local tools and HTTP/SSE endpoints for remote clients. Start the web server to enable Streamable HTTP at /mcp and SSE at /mcp-sse, then connect your MCP client to the appropriate endpoint.

How to install

Prerequisites: install Python 3.11 or newer and a runtime suitable for your platform. You also need a compatible package manager for your preferred installation flow.

# Prerequisites
# Ensure Python 3.11+ is installed
# Install the uv tool for global tool management (if you plan to install via uv)
python3 -m pip install uv

# Alternatively, install via pip for a Python environment
pip install z-image-studio

# If you want to run from source, clone the project and install in editable mode
# git clone https://github.com/iconben/z-image-studio.git
# cd z-image-studio
# pip install -e .

Notes on MCP setup and startup

The MCP server can be run in stdio mode for local agents or via the HTTP/SSE endpoints when you start the web server. Use the following entry points to start the MCP server locally.

Additional configuration and controls

You can control how the server starts, which transports are enabled, and how it exposes its tools. By default, you will have both streamable HTTP and SSE endpoints when the web server is running. If you need to disable all MCP endpoints, you can start the server with the appropriate flag to turn off MCP support.

Available tools

generate

Generate an image from a prompt using the MCP server, returning a structured response that includes metadata, a resource link, and a thumbnail preview.

list_models

List available image generation models and their details, including recommended models for your hardware.

list_history

Query the history of past generations, including prompts, seeds, and results.