depu-img-mcp
Enables text-only language models to understand images by forwarding image and prompt requests to vision model backends via MCP, returning descriptive text. Supports multiple OpenAI-compatible providers with task routing and safety checks.
README
depu-img-mcp
图像理解 MCP 服务器 —— 让不支持图像的文本模型通过 MCP 把图片 + prompt 转发给视觉模型,返回描述文本。自身不做推理,是协议适配与转发层。
特性
- MCP v2(
MCPServer+ Streamable HTTP,2026-07-28 规范),同时支持 stdio / SSE - Docker 部署(多阶段构建 + docker-compose)
- 多后端 provider:任意 OpenAI 兼容端点(LiteLLM 网关 / vLLM / OpenAI / DashScope / 智谱 / OpenRouter / Ollama …)
- 双层 prompt 注入:全局
base_vision_prompt+ per-callprompt+task_type路由(auto/general/ocr/ui/debug/describe) - 安全:magic byte 校验、大小限制、SSRF 防护、路径白名单
- 轻量 Web 后台(
/admin):查看配置 + 在线测试图片 - 配置双源:环境变量覆盖
config.toml,密钥支持${ENV}插值
快速开始
# 1. 准备 .env(至少填 ADMIN_TOKEN 和视觉后端 API key)
cp .env.example .env
# 编辑 .env 填入 API key
# 2. Docker 启动(首次会自动生成 ./conf/config.toml 默认配置)
docker compose up -d
# 3. 访问后台改配置(或直接编辑 ./conf/config.toml 后重启)
open http://localhost:8080/admin
# MCP 端点:http://localhost:8080/mcp
首次启动无需预先准备
config.toml:容器挂载./conf/目录,若里面没有配置文件,会自动从内置模板生成一份。之后可在/admin后台在线编辑,或直接改./conf/config.toml后docker compose restart。
本地运行(开发)
uv sync
uv run python -m depu_img_mcp # 默认 streamable-http
MCP_TRANSPORT=stdio uv run python -m depu_img_mcp # stdio 模式给 Claude Desktop 等
MCP 工具
image_understand
image_understand(
image: str, # http(s) URL 或 data:image/...;base64,... URI(内联 base64)。不支持本地文件路径(Docker 部署,容器看不到客户端文件系统)
prompt: str = "", # 用户问题;空则给出通用描述(物体、文字/OCR、布局、颜色等)
) -> str
provider / model 由后台 admin 统一配置,客户端不能指定。这用于企业内网服务,后端路由由运维通过
/admin后台决定。
只有一个工具。客户端传图 + prompt,服务器用后台配置的默认 provider 及其 model 进行视觉理解,返回文本描述。
配置示例
[[providers]]
name = "default"
type = "openai-compat"
base_url = "https://gateway.ai.depu.school/v1"
api_key = "${DEPU_GATEWAY_API_KEY}"
model = "Kimi-K2.7-Code"
auth_header = "bearer"
客户端配置(Claude Desktop 示例)
stdio 模式:
{
"mcpServers": {
"depu-img": {
"command": "python",
"args": ["-m", "depu_img_mcp"],
"env": { "MCP_TRANSPORT": "stdio", "DEPU_GATEWAY_API_KEY": "sk-..." }
}
}
}
HTTP 模式(支持远程 MCP 的客户端):
{
"mcpServers": {
"depu-img": { "url": "http://localhost:8080/mcp" }
}
}
License
MIT
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.