# mcp-image — MCP Image Generation & Editing Server

> mcp-image is an MCP server for image generation/editing with quality presets; verified 110★ and documents `npx -y mcp-image` configs for Cursor and Claude.

## Install

Merge the JSON below into your `.mcp.json`:

## Quick Use

```bash
# Cursor ~/.cursor/mcp.json (README example uses npx):
# {"mcpServers":{"mcp-image":{"command":"npx","args":["-y","mcp-image"],"env":{"GEMINI_API_KEY":"...","IMAGE_OUTPUT_DIR":"/abs/path"}}}}
# Or Claude Code:
claude mcp add mcp-image --scope user --env GEMINI_API_KEY=your-api-key --env IMAGE_OUTPUT_DIR=/absolute/path/to/images -- npx -y mcp-image
```

## Intro

mcp-image is an MCP server for image generation/editing with quality presets; verified 110★ and documents `npx -y mcp-image` configs for Cursor and Claude.

**Best for:** Teams who want a reusable MCP image tool with per-request quality controls and saved outputs

**Works with:** Any MCP client (Cursor/Codex/Claude Code) + Gemini API key or OpenAI API key (provider switch)

**Setup time:** 6-15 minutes

### Key facts (verified)

- GitHub: 110 stars · 19 forks · pushed 2026-05-13.
- License: MIT · owner avatar + repo URL verified via GitHub API.
- README-backed entrypoint: `npx -y mcp-image`.

## Main

- Treat output paths as part of the contract: set `IMAGE_OUTPUT_DIR` to an absolute path so every generated image is saved and reviewable.

- Pick a quality preset for your workflow: use fast for iteration, balanced for most work, and quality only for final deliverables (README preset table).

- If you use OpenAI mode, keep provider env vars isolated and never commit keys; test with a single prompt first to verify permissions and output formats.

### Source-backed notes

- README documents MCP configs for Codex (`~/.codex/config.toml`) and Cursor (`~/.cursor/mcp.json`) using `npx -y mcp-image`.
- README lists quality presets with a speed hint for `fast` (~30–40s) and env var control via `IMAGE_QUALITY`.
- README includes a Claude Code setup command (`claude mcp add ... -- npx -y mcp-image`) and notes `IMAGE_OUTPUT_DIR` must be an absolute path.

### FAQ

- **Do I need OpenAI to use it?**: No — README says Gemini is the default provider; OpenAI is optional via `IMAGE_PROVIDER=openai`.
- **Where do images go?**: To `IMAGE_OUTPUT_DIR` (absolute path) per README; the server creates the directory if missing.
- **How do I control speed vs quality?**: Use presets via `IMAGE_QUALITY` (fast/balanced/quality) as documented in README.

## Source & Thanks

> Source: https://github.com/shinpr/mcp-image
> License: MIT
> GitHub stars: 110 · forks: 19

---

<!-- ZH -->

## Quick Use

```bash
# Cursor ~/.cursor/mcp.json (README example uses npx):
# {"mcpServers":{"mcp-image":{"command":"npx","args":["-y","mcp-image"],"env":{"GEMINI_API_KEY":"...","IMAGE_OUTPUT_DIR":"/abs/path"}}}}
# Or Claude Code:
claude mcp add mcp-image --scope user --env GEMINI_API_KEY=your-api-key --env IMAGE_OUTPUT_DIR=/absolute/path/to/images -- npx -y mcp-image
```

## Intro

mcp-image 是用于图像生成/编辑的 MCP server，提供质量预设并可切换 provider；已验证 110★，并给出 Cursor/Codex/Claude Code 的 `npx -y mcp-image` 配置示例。

**Best for:** 希望把图像能力做成可复用 MCP 工具，并可控质量与落盘输出的团队

**Works with:** 任意 MCP 客户端（Cursor/Codex/Claude Code）+ Gemini 或 OpenAI API key（可切换 provider）

**Setup time:** 6-15 minutes

### Key facts (verified)

- GitHub：110 stars · 19 forks；最近更新 2026-05-13。
- 许可证：MIT；作者头像与仓库链接均已通过 GitHub API 复核。
- README 中可对照的入口命令：`npx -y mcp-image`。

## Main

- 把输出路径当作契约：设置 `IMAGE_OUTPUT_DIR` 为绝对路径，让每张生成图都能落盘可追溯。

- 按工作流选择质量档位：fast 用于迭代、balanced 用于常规、quality 只在最终交付时启用（README 表格）。

- 使用 OpenAI 模式时，把 provider 环境变量隔离并避免提交 key；先用一个最小 prompt 验证权限与输出格式。

### Source-backed notes

- README 给出 Codex（`~/.codex/config.toml`）与 Cursor（`~/.cursor/mcp.json`）的 MCP 配置示例，命令为 `npx -y mcp-image`。
- README 列出质量预设，并给出 `fast` 的速度提示（约 30–40 秒），且可通过 `IMAGE_QUALITY` 控制。
- README 提供 Claude Code 的安装命令（`claude mcp add ... -- npx -y mcp-image`），并强调 `IMAGE_OUTPUT_DIR` 必须为绝对路径。

### FAQ

- **必须用 OpenAI 吗？**：不必。README 说明默认是 Gemini；OpenAI 仅作为可选 provider（`IMAGE_PROVIDER=openai`）。
- **图片会保存到哪里？**：按 README：保存到 `IMAGE_OUTPUT_DIR`（绝对路径）；目录不存在会自动创建。
- **如何控制速度与质量？**：按 README：通过 `IMAGE_QUALITY` 选择 fast/balanced/quality。

## Source & Thanks

> Source: https://github.com/shinpr/mcp-image
> License: MIT
> GitHub stars: 110 · forks: 19


---
Source: https://tokrepo.com/en/workflows/mcp-image-mcp-image-generation-editing-server
Author: MCP Hub