{"$schema":"https://wellknown.network/schemas/agent-record-v1.json","schemaVersion":"1","id":"ag_c8w8ycvpvss4","handle":"visual-understand-mcp","url":"https://wellknown.network/agents/visual-understand-mcp","links":{"self":"https://wellknown.network/agents/visual-understand-mcp/record.json","html":"https://wellknown.network/agents/visual-understand-mcp","markdown":"https://wellknown.network/agents/visual-understand-mcp/record.md","api":"https://wellknown.network/api/v1/agents/visual-understand-mcp","status":"https://wellknown.network/api/v1/agents/visual-understand-mcp/status","claim":"https://wellknown.network/agents/visual-understand-mcp/claim","claimApi":"https://wellknown.network/api/v1/claims","claimDescriptor":"https://wellknown.network/agents/visual-understand-mcp/claim.json","badge":"https://wellknown.network/agents/visual-understand-mcp/badge.svg","openapi":"https://wellknown.network/openapi.json","history":"https://wellknown.network/api/v1/agents/visual-understand-mcp/history","tools":"https://wellknown.network/api/v1/agents/visual-understand-mcp/tools"},"ard":{"identifier":"urn:air::server:visual-understand-mcp","type":"application/mcp-server-card+json"},"kind":"mcp_server","declared":{"name":"visual-understand-mcp","summary":"视觉理解 MCP Server - 为不支持视觉的编码模型提供图片理解能力","description":"# 视觉理解 MCP Server\n\n为不支持视觉的编码模型提供图片理解能力。\n\n## 原理\n\n编码模型（如 GLM-5.1）遇到图片时，自动调用 MCP 工具 `understand_image`，由视觉模型完成图片识别，结果以文本回传给编码模型继续推理。无需切换模型，上下文不中断。\n\n```\n用户粘贴截图 → 编码模型调用 understand_image → 视觉模型识别 → 文字描述回传 → 继续编码\n```\n\n## 安装方式一：通过 PyPI / uvx 使用（推荐）\n\nMCP 配置示例：\n\n```json\n{\n  \"mcpServers\": {\n    \"visual-understand\": {\n      \"command\": \"uvx\",\n      \"args\": [\"visual-understand-mcp\"],\n      \"env\": {\n        \"VISION_API_BASE\": \"https://dashscope.aliyuncs.com/compatible-mode/v1\",\n        \"VISION_MODEL\": \"qwen-vl-max\",\n        \"VISION_API_KEY\": \"sk-xxx\"\n      }\n    }\n  }\n}\n```\n\n## 安装方式二：本地源码运行\n\n```json\n{\n  \"mcpServers\": {\n    \"visual-understand\": {\n      \"command\": \"uv\",\n      \"args\": [\n        \"--directory\", \"/path/to/visual-understand-mcp\",\n        \"run\", \"visual-understand-mcp\"\n      ],\n      \"env\": {\n        \"VISION_API_BASE\": \"https://dashscope.aliyuncs.com/compatible-mode/v1\",\n        \"VISION_MODEL\": \"qwen-vl-max\",\n        \"VISION_API_KEY\": \"sk-xxx\"\n      }\n    }\n  }\n}\n```\n\n## 配置项\n\n| 变量 | 说明 | 默认值 |\n|------|------|--------|\n| `VISION_API_BASE` | 视觉模型 API 地址 | 无（必填） |\n| `VISION_MODEL` | 视觉模型名称 | 无（必填） |\n| `VISION_API_KEY` | 视觉模型密钥 | 无（必填） |\n| `VISION_TEMPERATURE` | 输出随机性 | `0.1` |\n| `VISION_MAX_TOKENS` | 最大输出长度 | `12000` |\n| `VISION_TIMEOUT` | 请求超时（秒） | `120` |\n| `VISION_SYSTEM_PROMPT` | 视觉模型系统提示词 | 见下方 |\n\n前三个必填，其余可选。配置也可写入 `~/.visual-understand-mcp/config.json`，env 优先级更高。\n\n默认系统提示词：\n\n> 你是一个图片分析助手，仅用于理解图片内容。请按以下结构分析图片，根据实际内容调整详细程度：1. **文字内容** — 逐字提取图中所有可见文字，保留原始格式和层级关系。2. **错误信息与代码** — 如果图中包含错误信息或代码片段，必须原样输出，不做任何改写或概括；如果没有则忽略此项。3. **视觉布局与元素** — 描述空间排列、尺寸、颜色及关键元素间的关系。UI 截图请识别组件类型及其状态。4. **数据与指标** — 图表、表格请提取数值、坐标轴、标签、趋势及异常数据点。5. **整体概述** — 概括图片的主题、场景和关键信息，提供完整的上下文理解。不适用的部分跳过。只描述图片中可见的内容，不做推理、猜测或延伸解读。保持精确客观，不推测不可见的内容。\n\n## 建议写入 CLAUDE.md\n\n```markdown\n进行图片识别任务时，使用 visual-understand MCP 的 understand_image 工具。\n```\n\n## 本地调试\n\n```bash\nuv run mcp dev src/visual_understand_mcp/server.py\n```\n\n## License\n\nMIT","publisher":{"name":"xc","url":null},"homepage":null,"repository":null,"version":"0.1.2","license":null,"protocols":["mcp"],"tags":["image-recognition","mcp","vision","visual-understanding"],"pricing":null,"endpoints":[{"url":"pypi:visual-understand-mcp","type":"package_pypi","auth":null,"probeable":false}],"skills":null,"tools":null,"extra":null,"attribution":{"kind":"pypi","name":"pypi","summary":"pypi","version":"pypi","description":"pypi","publisherName":"pypi"}},"derived":{"capabilities":[{"slug":"media.image-understanding","name":"Image Understanding","confidence":1,"provenance":"declared"}],"categories":["media"],"language":"fr"},"observed":{"status":"unknown","statusReason":"Distributed as a package to run locally; no network endpoint to check.","lastOkAt":null,"lastProbedAt":null,"statusComputedAt":null,"reliability30d":null,"latestObservations":[],"tools":null,"package":{"name":"visual-understand-mcp","registry":"pypi","observedAt":"2026-10-22T23:26:17.579Z","publishedAt":"2026-06-24T11:52:37.204146Z","latestVersion":"0.1.2"},"toolSurface":null,"endpointFacts":[]},"verification":{"claimed":false,"claimedAt":null,"proofs":[]},"provenance":{"sources":[{"source":"pypi","key":"visual-understand-mcp","url":"https://pypi.org/project/visual-understand-mcp/","firstSeenAt":"2026-09-10T15:23:43.868Z","fetchedAt":"2026-10-22T23:24:44.826Z","normalizedAt":"2026-10-22T23:24:44.826Z"}]},"firstSeenAt":"2026-09-10T15:23:43.868Z","updatedAt":"2026-10-22T23:26:17.579Z"}