为纯文本大模型提供视觉能力的 MCP Server:读图 · 识字 · 解表格 · 复刻 UI。Give text-only LLMs vision via MCP: image description, OCR, VQA, UI layout & tables.
Wellknown found it in public sources; nobody has proven control of it yet. Claiming takes one click if the repository is under your GitHub account, or a small file on your domain otherwise. Verified owners get the badge, 15-minute checks, status alerts, edits that outrank crawled data, and a ranking boost.
Agents can do it too: POST https://wellknown.network/api/v1/claims with {"agent":"openeye-mcp","method":"well_known_file"} — machine-readable steps at claim.json, guide at /docs/claim.
Everything here was measured by our prober or read from a registry. Nothing is self-reported.
Attributed to the source that supplied each field. Treated as claims, not facts.
<div align="center"> # OpenEye **为纯文本大模型提供视觉能力的 MCP Server —— 让不会看图的模型也能读图、识字、解表格、复刻 UI。** <a href="https://pypi.org/project/openeye-mcp/"><img src="https://img.shields.io/pypi/v/openeye-mcp.svg" alt="PyPI version"></a> <img src="https://img.shields.io/pypi/pyversions/openeye-mcp.svg" alt="Python 3.11+"> <a href="LICENSE"><img src="https://img.shields.io/badge/License-MIT-green.svg" alt="MIT License"></a> <img src="https://img.shields.io/badge/transport-stdio-blueviolet.svg" alt="MCP stdio transport"> </div> OpenEye 是一个 stdio 类型的 [MCP](https://modelcontextprotocol.io) 服务器:把视觉模型(GPT / Claude / Gemini / 通义 / 智谱 / 本地 Ollama 等)包装成 6 个 MCP 工具,接进 Claude Code、Cursor、Cline 等 coding agent,原本只能处理文本的模型就获得了识图能力——分析 UI 截图、照抄报错文字、读表格、按图还原前端。 ## 功能特性 - **6 个识图工具**:图像描述、OCR 取字、视觉问答、UI 布局结构化分析、表格转 Markdown、多图批量分析 + 跨图汇总 - **5 类视觉后端**:OpenAI 兼容协议(默认)/ Anthropic Messages / OpenAI Responses / Gemini generateContent / Gemini Interactions,一个 `VISION_PROVIDER` 切换 - **任意 OpenAI 兼容厂商**:改 `OPENAI_BASE_URL` 就换厂商——通义、智谱、阶跃星辰、Kimi、DeepSeek、本地 Ollama / vLLM 都能直接接 - **三种图片来源**:本地路径、`http(s)` 公网 URL、`data:image/...;base64,` data URI,同一入口自动识别 - **结构化输出**:布局与表格走 JSON 模式,端点不支持 `response_format` 时自动降级并用容错解析兜底 - **省 token 与提速**:图片等比压缩(JPEG 化)、结果缓存(LRU + TTL)、已解析图片缓存,重复问同一张图不再重复上传 - **进度上报**:支持 MCP `progressToken` 的客户端能看到「正在请求视觉后端(第 N 次)」这类实时进度 - **默认安全的取图链路**:SSRF 防护(DNS 固定 + 逐跳重定向校验)、图片大小上限、凭据脱敏、可选本地目录白名单、单次调用总时间预算 - **零 SDK 依赖**:只用 `httpx` 手写协议,不装各厂商 SDK ## 安装 需要 Python 3.11+、一个支持图片输入的视觉模型 API Key(纯文本模型无法看图),以及任一支持 MCP 的客户端(Claude Code / Cursor / Cline / Windsurf / Codex CLI / Zed / Continue / Roo Code / opencode 等)。 ```bash pip install openeye-mcp ``` **虚拟环境安装**(推荐;客户端配置里用该环境的绝对路径): ```bash python -m venv .venv source .venv/bin/activate # Windows: .venv\Scripts\activate pip install openeye-mcp ``` **从源码安装**(开发与贡献): ```bash git clone https://github.com/ouli-1242/openeye-mcp.git cd openeye-mcp pip install -e ".[dev]" # edit…
Mapped onto the structured taxonomy from declared text and observed tool names. Confidence shown for derived entries.
Every source is kept verbatim. Field changes are logged as events.