gemini-image-mcp
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@gemini-image-mcp generate a futuristic city skyline at sunset"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
gemini-image-mcp
一个基于 MCP (Model Context Protocol) 的 Node.js 服务:通过本地 OpenAI-compatible 网关(默认 http://127.0.0.1:8317)调用 Gemini 图片模型 gemini-3-pro-image-preview 生成图片,供 Claude Code/Claude Desktop 等 IDE 工具动态使用。
1) 安装
npm installRelated MCP server: Gemini Gen MCP
2) 环境变量(可选)
OPENAI_BASE_URL:OpenAI-compatible 服务地址(默认http://127.0.0.1:8317)OPENAI_API_KEY:如果你的网关需要鉴权就填(可留空)OPENAI_MODEL:默认gemini-3-pro-image-previewOPENAI_IMAGE_SIZE:可选,仅作为未传入size时的默认值;建议让客户端在调用generate_image时自己传sizeOPENAI_IMAGE_MODE:chat|images|auto,默认chat(CLIProxyAPI 这类网关通常用/v1/chat/completions出图;若你的网关支持/v1/images/generations可设为images)OPENAI_IMAGE_RETURN:path|image,默认path(path会把图片保存到本地并返回文件路径,避免 base64 导致 token 暴涨;image返回 MCPimagecontent)OPENAI_IMAGE_OUT_DIR:保存目录(默认debug-output/;相对路径以项目根目录为基准)OPENAI_DEBUG:设为1时会在 stderr 打印上游请求信息(不打印 key)OPENAI_TIMEOUT_MS:默认120000
可参考 .env.example。
3) 本地调试(不用放进 Claude Code)
推荐把 .env.example 复制成 .env,然后在 .env 里填好 OPENAI_API_KEY(.gitignore 已忽略 .env)。
直连上游调试(确认你的
http://127.0.0.1:8317是否能出图):npm run debug:upstream -- --prompt "A beautiful sunset over mountains" --size 1024x1024
走 MCP 工具调试(等价于 Claude Code 调用
generate_image):npm run debug:mcp -- --prompt "A beautiful sunset over mountains" --n 1 --size 1024x1024
图片会输出到 debug-output/。
4) 作为 MCP Server 使用(stdio)
该项目是 stdio 传输方式的 MCP Server,不建议直接在终端手动运行(会等待客户端请求)。
在 Claude Code / Claude Desktop 的 MCP 配置里添加类似如下(按你的实际路径修改):
{
"mcpServers": {
"gemini-image": {
"command": "node",
"args": ["d:/task/myself/nodejs/geminiimagemcp/src/index.js"],
"env": {
"OPENAI_BASE_URL": "http://127.0.0.1:8317",
"OPENAI_API_KEY": "<YOUR_KEY>",
"OPENAI_MODEL": "gemini-3-pro-image-preview"
}
}
}
}也可以直接参考 mcp.example.json。
5) 可用工具
generate_image入参:
prompt(必填),size(可选),n(可选,1-4),output(可选:path|image),outDir(可选)返回:默认返回保存后的图片文件路径(多行);
output=image时返回 MCPimagecontent(base64 + mimeType)
Available Tools
1 toolgenerate_imageA
通过 OpenAI-compatible 接口调用 Gemini 的 "gemini-3-pro-image-preview" 生成图片:默认保存到本地并返回文件路径(避免 base64 导致 token 暴涨);也可用 output=image 返回 MCP image content。
Inputs: prompt (required), size, n, output, +1 more.
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
TDQS
Scored across 1 tool
With only one tool, there is no possibility of confusion between tools. The single tool has a clearly defined purpose.
Only one tool exists, so naming conventions are inherently consistent. The name 'generate_image' is clear and verbs are used appropriately.
The server focuses on image generation, and one tool is reasonable for this narrow scope. While it is minimal, it fully covers the core functionality without unnecessary complexity.
The tool covers all essential aspects of image generation: it offers both local file saving and MCP image content output. No obvious gaps for the intended purpose.
Maintenance
Related MCP Connectors
MCP server for OpenAI API (chat completions, image generation, embeddings) via AceDataCloud
MCP server for Qwen Image 3 AI image generation
Generate AI images and videos from any compatible MCP client.
MCP server for Flux AI image generation
Related MCP Servers
- AlicenseBqualityDmaintenanceAn MCP server for image generation using the Gemini API.1126 npm2MIT
- AlicenseBqualityDmaintenanceMCP server for generating images and audio using Google's Gemini AI models.272 PyPI2MIT
- AlicenseAqualityCmaintenanceMCP server for Google Gemini image generation with configurable model support, enabling text-to-image generation, image editing, and iterative refinement.640MIT
- FlicenseNot gradedqualityDmaintenanceAn MCP server for AI-powered image processing (generate, edit, vary, analyze) supporting OpenAI, Gemini, Ideogram, and custom relay endpoints.-