NovelAI Image MCP

An MCP (Model Context Protocol) server that exposes NovelAI image generation as tools for AI agents — Claude Desktop, Cline, custom agents, and remote clients.

Built on FastMCP 4 (the fastmcp framework over the MCP SDK v2 mcp>=2.0.0), it lets an agent generate images (txt2img / img2img / inpaint), upscale, run Director tools (line art, emotion, background removal, …), annotate with ControlNet, suggest tags, encode vibes, and query account subscription — all through the standard MCP tool interface.

---

Highlights

• 11 MCP tools covering the full NovelAI image API surface.

• Two transports: stdio (local agents) and streamable-http (remote / multi-client).

• Dual return shape: base64 Image content blocks (the agent sees the image) and PNG saved to disk (path returned as text).

• Async + sync: async tool handlers + a typer CLI for direct invocation.

• Agent skills: three skills.sh packages (novelai-cli, novelai-mcp-tools, novelai-workflows) that teach AI agents how to drive the CLI and MCP tools.

• uv-managed monorepo, MIT-licensed, Docker-ready, GitHub Pages docs.

---

Get started

---

Quick links

• Quick start — install, configure, and generate your first image

• Tools reference — every MCP tool, parameter, and example

• Transports — stdio vs streamable-http

• Agent skills — skills.sh packages that teach AI agents the CLI + MCP tools

• Agent host setup — Claude Desktop, Cline, Cursor, Continue, Windsurf, Codex CLI

• API reference — autodoc-generated Python API

• Contributing — how to hack on the server

---

Project links

• Source: https://github.com/xinvxueyuan/NovelAI-Image-MCP

• Issues: https://github.com/xinvxueyuan/NovelAI-Image-MCP/issues

• License: MIT

• Changelog: Keep a Changelog

• NovelAI API docs: https://image.novelai.net/docs/index.html