{
 "jsonrpc": "2.0",
 "id": 2,
 "result": {
  "tools": [
   {
    "name": "lanhu_resolve_invite_link",
    "description": "Resolve Lanhu invite/share link to actual project URL\n\nUSE THIS WHEN: User provides invite link (lanhuapp.com/link/#/invite?sid=xxx)\n\nPurpose: Convert invite link to usable project URL with tid/pid/docId parameters\n\nReturns:\n    Resolved URL and parsed parameters",
    "inputSchema": {
     "additionalProperties": false,
     "properties": {
      "invite_url": {
       "description": "Lanhu invite link. Example: https://lanhuapp.com/link/#/invite?sid=xxx",
       "type": "string"
      }
     },
     "required": [
      "invite_url"
     ],
     "type": "object"
    },
    "outputSchema": {
     "additionalProperties": true,
     "type": "object"
    },
    "_meta": {
     "fastmcp": {
      "tags": []
     }
    }
   },
   {
    "name": "lanhu_list_product_documents",
    "description": "[PRD/Requirement Document Discovery] List all product documents (PRD/prototype) in a Lanhu project.\n\nUSE THIS WHEN user says: 有哪些需求文档, 列出项目的文档, 项目下的PRD列表, 产品文档列表,\n    这个项目有什么文档, 文档列表, list documents, product_documents, 所有文档, 全部 PRD\nDO NOT USE for: 获取某个具体文档的页面 (use lanhu_get_pages instead),\n    UI设计图列表 (use lanhu_get_designs instead)\n\nPurpose: Discover available PRD/requirement documents in a project so the user can pick\n    the right one to analyze. Typically used BEFORE lanhu_get_pages when docId is unknown.\n\nReturns a dict with top-level metadata and a simplified `documents` list. Each document\ncarries a ready-to-use `doc_url` that can be fed straight into lanhu_get_pages.",
    "inputSchema": {
     "additionalProperties": false,
     "properties": {
      "url": {
       "description": "Lanhu project URL. Example: https://lanhuapp.com/web/#/item/project/product?tid=xxx&pid=xxx (docId optional, will be ignored). Required params: tid, pid. If you have an invite link, use lanhu_resolve_invite_link first!",
       "type": "string"
      }
     },
     "required": [
      "url"
     ],
     "type": "object"
    },
    "outputSchema": {
     "additionalProperties": true,
     "type": "object"
    },
    "_meta": {
     "fastmcp": {
      "tags": []
     }
    }
   },
   {
    "name": "lanhu_get_pages",
    "description": "[PRD/Requirement Document] Get page list of Lanhu Axure prototype - CALL THIS FIRST before analyzing\n\nUSE THIS WHEN user says: 需求文档, 需求, PRD, 产品文档, 原型, 交互稿, Axure, 看看需求, 帮我看需求, 需求分析\nDO NOT USE for: UI设计图, 设计稿, 视觉设计, 切图 (use lanhu_get_designs instead)\n\nPurpose: Get page list of PRD/requirement/prototype document. Must call this BEFORE lanhu_get_ai_analyze_page_result.\n\nReturns:\n    Page list and document metadata",
    "inputSchema": {
     "additionalProperties": false,
     "properties": {
      "url": {
       "description": "Lanhu URL with docId parameter (indicates PRD/prototype document). Example: https://lanhuapp.com/web/#/item/project/product?tid=xxx&pid=xxx&docId=xxx. Required param: pid. tid and docId recommended. Supports detailDetach format: ?pid=xxx&image_id=xxx. If you have an invite link, use lanhu_resolve_invite_link first!",
       "type": "string"
      }
     },
     "required": [
      "url"
     ],
     "type": "object"
    },
    "outputSchema": {
     "additionalProperties": true,
     "type": "object"
    },
    "_meta": {
     "fastmcp": {
      "tags": []
     }
    }
   },
   {
    "name": "lanhu_get_ai_analyze_page_result",
    "description": "[PRD/Requirement Document] Analyze Lanhu Axure prototype pages - GET VISUAL CONTENT\n\nUSE THIS WHEN user says: 需求文档, 需求, PRD, 产品文档, 原型, 交互稿, Axure, 看看需求, 帮我看需求, 分析需求, 需求分析\nDO NOT USE for: UI设计图, 设计稿, 视觉设计, 切图 (use lanhu_get_ai_analyze_design_result instead)\n\nFOUR-STAGE WORKFLOW (ZERO OMISSION):\n1. STAGE 1: Call with mode=\"text_only\" and page_names=\"all\" for global text scan\n   - Purpose: Build god's view, understand structure, design grouping strategy\n   - Output: Text only (fast)\n   - ⚠️ IMPORTANT: After STAGE 1, MUST ask user to choose analysis_mode!\n\n2. STAGE 2: Call with mode=\"full\" for each group (output format varies by analysis_mode)\n   - developer: Extract ALL details (fields, rules, flows) - for coding\n   - tester: Extract test scenarios, validation points, field rules - for test cases\n   - explorer: Extract core functions only (3-5 points) - for requirement review\n\n3. STAGE 3: Reverse validation (format varies by analysis_mode)\n\n4. STAGE 4: Generate deliverable (format varies by analysis_mode)\n   - developer: Detailed requirement doc + global flowchart\n   - tester: Test plan + test case list + field validation table\n   - explorer: Review PPT-style doc + module table + dependency diagram\n\nReturns:\n    - mode=\"text_only\": Text content only (for fast global scan)\n    - mode=\"full\": Visual + text + design style info (format determined by analysis_mode)\n      Each page includes [设计样式参考] with:\n        - 文字颜色: exact text colors used (rgba/rgb values, sorted by frequency)\n        - 背景颜色: exact background colors used\n        - 字体规格: font-size / font-weight / color combinations\n        - 页面图片资源: all images used on the page with dimensions and local paths\n      When generating code, you MUST use these exact color/font/size values from\n      [设计样式参考] instead of guessing. For images, use the local file paths provided.",
    "inputSchema": {
     "additionalProperties": false,
     "properties": {
      "url": {
       "description": "Lanhu URL with docId parameter (indicates PRD/prototype document). Example: https://lanhuapp.com/web/#/item/project/product?tid=xxx&pid=xxx&docId=xxx. Required param: pid. tid and docId recommended. Supports detailDetach format. If you have an invite link, use lanhu_resolve_invite_link first!",
       "type": "string"
      },
      "page_names": {
       "anyOf": [
        {
         "type": "string"
        },
        {
         "items": {
          "type": "string"
         },
         "type": "array"
        }
       ],
       "description": "Page name(s) to analyze. Use 'all' for all pages, single name like '退款流程', or list like ['退款流程', '用户中心']. Get exact names from lanhu_get_pages first!"
      },
      "mode": {
       "default": "full",
       "description": "Analysis mode: 'text_only' (fast global scan, text only for overview) or 'full' (detailed analysis with images+text). Default: 'full'",
       "type": "string"
      },
      "analysis_mode": {
       "default": "developer",
       "description": "Analysis perspective (MUST be chosen by user after STAGE 1): 'developer' (detailed for coding), 'tester' (test scenarios/validation), 'explorer' (quick overview for review). Default: 'developer'",
       "type": "string"
      }
     },
     "required": [
      "url",
      "page_names"
     ],
     "type": "object"
    },
    "_meta": {
     "fastmcp": {
      "tags": []
     }
    }
   },
   {
    "name": "lanhu_get_designs",
    "description": "[UI Design] Get Lanhu UI design image list - CALL THIS FIRST before analyzing designs\n\nUSE THIS WHEN user says: UI设计图, 设计图, 设计稿, 视觉设计, UI稿, 看看设计, 帮我看设计图, 设计评审\nDO NOT USE for: 需求文档, PRD, 原型, 交互稿, Axure (use lanhu_get_pages instead)\nDO NOT USE for: 切图, 图标, 素材 (use lanhu_get_design_slices instead)\n\nPurpose: Get list of UI design images from designers. Must call this BEFORE lanhu_get_ai_analyze_design_result.\n\nReturns:\n    Design image list and project metadata",
    "inputSchema": {
     "additionalProperties": false,
     "properties": {
      "url": {
       "description": "Lanhu URL WITHOUT docId (indicates UI design project, not PRD). Example: https://lanhuapp.com/web/#/item/project/stage?tid=xxx&pid=xxx. Required param: pid. tid is optional. Supports detailDetach format: ?pid=xxx&image_id=xxx",
       "type": "string"
      }
     },
     "required": [
      "url"
     ],
     "type": "object"
    },
    "outputSchema": {
     "additionalProperties": true,
     "type": "object"
    },
    "_meta": {
     "fastmcp": {
      "tags": []
     }
    }
   },
   {
    "name": "lanhu_get_ai_analyze_design_result",
    "description": "[UI Design] Analyze Lanhu UI design images - GET VISUAL CONTENT + HTML CODE\n\nUSE THIS WHEN user says: UI设计图, 设计图, 设计稿, 视觉设计, UI稿, 看看设计, 帮我看设计图, 设计评审\nDO NOT USE for: 需求文档, PRD, 原型, 交互稿, Axure (use lanhu_get_ai_analyze_page_result instead)\nDO NOT USE for: 切图, 图标, 素材 (use lanhu_get_design_slices instead)\n\nWORKFLOW: First call lanhu_get_designs to get design list, then call this to analyze specific designs.\n\nReturns:\n    Visual representation of UI design images AND HTML+CSS code for each design.\n    First block: summary text with \"设计图 1/2/3...\" and each design's HTML code.\n    Following blocks: images in the same order as 设计图 1, 2, 3... (image N = design N).\n    \nCRITICAL - How to use the returned HTML+CSS (MUST follow this workflow):\n\n    ⚠️ AUTHORITY PRIORITY (highest → lowest):\n        1. HTML+CSS code  — the PRIMARY source of truth for all visual parameters\n        2. Design Tokens  — supplementary reference for gradients/borders/shadows\n        3. Design Image   — visual verification ONLY, never override CSS values\n\n    The returned HTML+CSS is the DESIGN SPECIFICATION generated from design schema.\n    Every CSS property value (color, size, spacing, font, gradient, border-radius,\n    etc.) is extracted from the original design data and MUST be used as-is.\n\n    RULE 1 - HTML+CSS IS DESIGN SPEC, COPY CSS VALUES DIRECTLY:\n        The CSS values are the single source of truth for all design parameters.\n        You MUST directly copy/reuse the exact CSS property values from the code.\n        DO NOT modify, simplify, or \"improve\" any CSS value. Specifically:\n          - DO NOT change rgba() to hex or vice versa (keep rgba(255,115,10,1) as-is)\n          - DO NOT round or simplify numbers (keep 0.30000001192092896 as-is)\n          - DO NOT replace linear-gradient with solid colors\n          - DO NOT change font-family order or remove fallback fonts\n          - DO NOT adjust margin/padding values for \"cleaner\" numbers\n          - DO NOT replace any img src or background-url with SVG, CSS shapes, or emoji\n          - DO NOT omit any visual element from the design\n        The HTML DOM structure and class names indicate layout intent (flex-row=Row,\n        flex-col=Column, justify-between=SpaceBetween, etc.), adapt them to the\n        target framework's component model while keeping all CSS values unchanged.\n\n    RULE 2 - DETECT USER PROJECT AND GENERATE FRAMEWORK-APPROPRIATE CODE:\n        STEP 1: Read project config files (package.json, tsconfig.json, pubspec.yaml,\n                build.gradle, Podfile, etc.) to detect framework and styling approach.\n        STEP 2: Generate code matching the detected framework:\n          - React/Next.js  → JSX component + CSS Modules / styled-components / Tailwind\n          - Vue/Nuxt       → Single File Component (.vue) with <style scoped>\n          - Angular        → component.ts + component.html + component.css\n          - Svelte         → Component.svelte with <style>\n          - Flutter        → StatelessWidget with EdgeInsets, BoxDecoration, etc.\n          - SwiftUI        → View struct with ViewModifier\n          - Android Compose→ @Composable function with Modifier\n          - Plain HTML     → Single self-contained .html file with inline <style>\n        STEP 3: Follow the project's existing conventions (file naming, directory\n                structure, styling approach). If no framework detected, default to\n                plain HTML single file.\n        CSS-to-platform property mapping reference:\n          width/height px    → Android: dp, iOS: pt, Flutter: logical pixels\n          font-size px       → Android: sp, iOS: pt, Flutter: fontSize\n          margin/padding     → Keep proportions, convert px to dp/pt\n          border-radius      → Android: dp, iOS: cornerRadius, Flutter: BorderRadius\n          color rgba()       → Android: Color.argb(), iOS: UIColor, Flutter: Color\n          linear-gradient    → Android: GradientDrawable, iOS: CAGradientLayer\n          flex-row / flex-col→ Row/Column (Flutter), HStack/VStack (SwiftUI)\n          position:absolute  → Stack+Positioned (Flutter), ZStack (SwiftUI)\n\n    RULE 3 - IMAGE ASSETS USE LOCAL PATHS (MANDATORY):\n        The returned HTML+CSS already uses LOCAL paths (./assets/slices/xxx.png)\n        for all image resources. A download mapping table is provided below each\n        design's HTML code, listing: local_path ← remote_download_url.\n        You MUST:\n          1. Download ALL images from the mapping table to the project's local\n             assets directory BEFORE generating final code.\n          2. Keep using local paths in the generated code. Adapt paths to the\n             target framework convention:\n               React/Vue   → import coverImg from '@/assets/slices/cover.png'\n               Flutter     → AssetImage('assets/images/cover.png')\n               Plain HTML  → <img src=\"./assets/slices/cover.png\">\n          3. NEVER use remote lanhu CDN URLs in any generated code.\n        Additionally, call lanhu_get_design_slices(url, design_name) to get the\n        full slice list for more fine-grained assets (icons, background images, etc.).\n\n    RULE 4 - CROSS-REFERENCE DESIGN TOKENS (SUPPLEMENTARY ONLY):\n        Design Tokens (if present) are extracted from the raw Sketch data.\n        They serve as SUPPLEMENTARY reference for properties that HTML+CSS may\n        not fully express (e.g. complex gradients, multi-stop fills, shadows).\n        Use Design Tokens to ENRICH the code, not to override HTML+CSS values.\n        Only when a CSS property is clearly MISSING (not just different) from the\n        HTML+CSS, use the Design Token value as a supplement.\n        Focus on: gradients, border styles, border-radius, opacity, shadows.\n\n    RULE 5 - POST-GENERATION FIDELITY AUDIT (MANDATORY, NEVER SKIP):\n        After generating code in ANY target platform/language (HTML/CSS, React,\n        Vue, Flutter, SwiftUI, Android Compose, etc.), perform a property-by-property\n        comparison against the design spec HTML+CSS. Map each CSS property to its\n        platform equivalent and verify the value is preserved exactly:\n          ① size constraint: fixed height in spec → must NOT become flexible/wrap\n              HTML: height not min-height | Flutter: fixed SizedBox, not Flexible\n              SwiftUI: .frame(height:) not omitted | Compose: height() not wrapContent\n          ② clipping: overflow:hidden in spec → must clip content in all platforms\n              HTML: overflow:hidden | Flutter: ClipRect/ClipRRect | SwiftUI: .clipped()\n              Compose: clip()/clipToBounds | Android: android:clipChildren=\"true\"\n          ③ color value: rgba(r,g,b,a) must be converted to platform format exactly\n              HTML: keep rgba() | Flutter: Color.fromRGBO() | SwiftUI: Color(red:green:blue:opacity:)\n              Compose: Color(r,g,b,a) | Android XML: #AARRGGBB — values must not drift\n          ④ gradient: linear-gradient must map to platform gradient, not solid color\n              Flutter: LinearGradient | SwiftUI: LinearGradient | Compose: Brush.linearGradient\n          ⑤ absolute positioning: left/top values must map to exact offsets\n              Flutter: Positioned(left:,top:) | SwiftUI: .offset() or .position()\n              Compose: Box+Modifier.offset() | HTML: position:absolute + left/top\n          ⑥ font: family, weight, size must all be preserved; fallback list for HTML\n          ⑦ spacing: every margin/padding direction value must be unchanged\n              HTML: margin/padding | Flutter: EdgeInsets | SwiftUI: .padding()\n              Compose: Modifier.padding() | Android: android:layout_margin / android:padding\n          ⑧ image assets: no image replaced by SVG/CSS shape/emoji/placeholder\n          ⑨ element completeness: every visible element in spec must appear in code\n          ⑩ no remote URLs: no lanhu CDN URLs in any generated asset path\n        For each difference found, state explicitly whether it is an intentional\n        platform adaptation (e.g. px→dp unit conversion) or an error (value changed).\n        All errors MUST be corrected before delivering the final code.\n\n    DESIGN IMAGE is for visual verification ONLY. It has the LOWEST priority.\n    NEVER use the design image to override any CSS value from the HTML+CSS code.",
    "inputSchema": {
     "additionalProperties": false,
     "properties": {
      "url": {
       "description": "Lanhu URL WITHOUT docId (indicates UI design project). Example: https://lanhuapp.com/web/#/item/project/stage?tid=xxx&pid=xxx. Required param: pid. tid is optional. Supports detailDetach format: ?pid=xxx&image_id=xxx",
       "type": "string"
      },
      "design_names": {
       "anyOf": [
        {
         "type": "string"
        },
        {
         "items": {
          "type": "string"
         },
         "type": "array"
        }
       ],
       "description": "Design name(s) or index number(s). 'all' = all designs. Number (e.g. 6) = the 6th item in lanhu_get_designs list (by 'index' field), NOT by name prefix. Exact name (e.g. '6_friend页_挂件墙') = match by full name. Get names/index from lanhu_get_designs first."
      }
     },
     "required": [
      "url",
      "design_names"
     ],
     "type": "object"
    },
    "_meta": {
     "fastmcp": {
      "tags": []
     }
    }
   },
   {
    "name": "lanhu_get_design_slices",
    "description": "[UI Slices/Assets] Get slice/asset info from Lanhu design for download\n\nUSE THIS WHEN user says: 切图, 下载切图, 图标, icon, 素材, 资源, 导出切图, 下载素材, 获取图标\nDO NOT USE for: 需求文档, PRD, 原型 (use lanhu_get_pages instead)\nDO NOT USE for: 看设计图, 设计评审 (use lanhu_get_designs instead)\n\nWORKFLOW: First call lanhu_get_designs to get design list, then call this to get slices from specific design.\n\nReturns:\n    Slice list with download URLs, AI will handle smart naming and batch download",
    "inputSchema": {
     "additionalProperties": false,
     "properties": {
      "url": {
       "description": "Lanhu URL WITHOUT docId (indicates UI design project). Example: https://lanhuapp.com/web/#/item/project/stage?tid=xxx&pid=xxx. Required param: pid. tid is optional. Supports detailDetach format: ?pid=xxx&image_id=xxx",
       "type": "string"
      },
      "design_name": {
       "description": "Exact design name (single design only, NOT 'all'). Example: '首页设计', '登录页'. Must match exactly with name from lanhu_get_designs result!",
       "type": "string"
      },
      "include_metadata": {
       "default": true,
       "description": "Include color, opacity, shadow info",
       "type": "boolean"
      }
     },
     "required": [
      "url",
      "design_name"
     ],
     "type": "object"
    },
    "outputSchema": {
     "additionalProperties": true,
     "type": "object"
    },
    "_meta": {
     "fastmcp": {
      "tags": []
     }
    }
   },
   {
    "name": "lanhu_say",
    "description": "Post message to team message board\n\nUSE THIS WHEN user says: 有话说, 留言, 发消息, 通知团队, 告诉xxx, @张三, @李四, 共享给xxx, 分享给xxx, 发给xxx, 写给xxx, 转发给xxx\n\nMessage type description:\n- normal: Normal message/notification (default)\n- task: Query task - Only for query operations (query code, query database, query TODO, etc.), NO code modification\n- question: Question message - Needs answer from others\n- urgent: Urgent message - Needs immediate attention\n- knowledge: Knowledge base - Long-term preserved experience, pitfalls, notes, best practices\n\nSecurity restrictions:\ntask type can only be used for query operations, including:\n- Query code location, code logic\n- Query database table structure, data\n- Query test methods, test coverage\n- Query TODO, comments\n- Forbidden: Modify code, delete files, execute commands, commit code\n\nKnowledge use cases:\n- Pitfalls encountered and solutions\n- Testing notes\n- Development experience and best practices\n- Common FAQ\n- Technical decision records\n\nPurpose: Post message to project message board, can @ specific person to send Feishu notification\n\nReturns:\n    Post result, including message ID and details",
    "inputSchema": {
     "additionalProperties": false,
     "properties": {
      "url": {
       "description": "蓝湖URL（含tid和pid）。例: https://lanhuapp.com/web/#/item/project/product?tid=xxx&pid=xxx&docId=xxx。会自动提取项目和文档信息",
       "type": "string"
      },
      "summary": {
       "description": "留言标题/概要",
       "type": "string"
      },
      "content": {
       "description": "留言详细内容",
       "type": "string"
      },
      "mentions": {
       "anyOf": [
        {
         "items": {
          "type": "string"
         },
         "type": "array"
        },
        {
         "type": "null"
        }
       ],
       "default": null,
       "description": "⚠️@提醒人名。必须是具体人名，例如: 张三/李四/王五/赵六等。禁止使用角色名(后端/前端等)！"
      },
      "message_type": {
       "anyOf": [
        {
         "type": "string"
        },
        {
         "type": "null"
        }
       ],
       "default": null,
       "description": "留言类型。可选: normal(普通留言), task(查询任务-仅限查询操作,禁止修改代码), question(需要回答的问题), urgent(紧急通知), knowledge(知识库-长期保存的经验知识)。默认: normal"
      }
     },
     "required": [
      "url",
      "summary",
      "content"
     ],
     "type": "object"
    },
    "outputSchema": {
     "additionalProperties": true,
     "type": "object"
    },
    "_meta": {
     "fastmcp": {
      "tags": []
     }
    }
   },
   {
    "name": "lanhu_say_list",
    "description": "Get message list with filtering and search\n\nUSE THIS WHEN user says: 查看留言, 有什么消息, 谁@我了, 留言列表, 消息列表\n\nSupports two modes:\n1. Provide specific URL: Query messages in that project\n2. url='all' or url=None: Query messages in all projects (global mode)\n\nImportant: To prevent AI context overflow, it is recommended:\n1. Use filter_type to filter by type\n2. Use search_regex for further filtering (regex, AI can generate itself)\n3. Use limit to limit the number of returned messages\n4. Unless user explicitly requests \"view all\", filters must be used\n\nExample:\n- Query all knowledge: filter_type=\"knowledge\"\n- Search containing \"test\" or \"refund\": search_regex=\"test|refund\"\n- Query tasks and containing \"database\": filter_type=\"task\", search_regex=\"database\"\n- Limit to 10 latest: limit=10\n\nPurpose: Get message board message summary list, supports type filtering, regex search and quantity limit\n\nReturns:\n    Message list, including mentions_me count",
    "inputSchema": {
     "additionalProperties": false,
     "properties": {
      "url": {
       "anyOf": [
        {
         "type": "string"
        },
        {
         "type": "null"
        }
       ],
       "default": null,
       "description": "蓝湖URL或'all'。不传或传'all'=查询所有项目；传具体URL=查询单个项目"
      },
      "filter_type": {
       "anyOf": [
        {
         "type": "string"
        },
        {
         "type": "null"
        }
       ],
       "default": null,
       "description": "筛选留言类型: normal/task/question/urgent/knowledge。不传则返回所有类型"
      },
      "search_regex": {
       "anyOf": [
        {
         "type": "string"
        },
        {
         "type": "null"
        }
       ],
       "default": null,
       "description": "正则表达式搜索（在summary和content中匹配）。例: '测试|退款|坑'。建议使用以避免返回过多消息"
      },
      "limit": {
       "default": null,
       "description": "限制返回消息数量（防止上下文爆炸）。不传则不限制",
       "title": "Limit"
      }
     },
     "type": "object"
    },
    "outputSchema": {
     "additionalProperties": true,
     "type": "object"
    },
    "_meta": {
     "fastmcp": {
      "tags": []
     }
    }
   },
   {
    "name": "lanhu_say_detail",
    "description": "Get message detail (supports batch query)\n\nUSE THIS WHEN user says: 查看详情, 看看内容, 详细内容, 消息详情\n\nTwo modes:\n1. Provide url: Parse project_id from url, query messages in that project\n2. url='all'/None + project_id: Global mode, need to manually specify project_id\n\nPurpose: Get full content of messages by message ID\n\nReturns:\n    Message detail list with full content",
    "inputSchema": {
     "additionalProperties": false,
     "properties": {
      "message_ids": {
       "description": "消息ID。单个数字或数组。例: 1 或 [1,2,3]"
      },
      "url": {
       "anyOf": [
        {
         "type": "string"
        },
        {
         "type": "null"
        }
       ],
       "default": null,
       "description": "蓝湖URL。传URL则自动解析项目ID；不传则需手动提供project_id参数"
      },
      "project_id": {
       "anyOf": [
        {
         "type": "string"
        },
        {
         "type": "null"
        }
       ],
       "default": null,
       "description": "项目ID。仅在不传url时需要，用于全局查询模式"
      }
     },
     "required": [
      "message_ids"
     ],
     "type": "object"
    },
    "outputSchema": {
     "additionalProperties": true,
     "type": "object"
    },
    "_meta": {
     "fastmcp": {
      "tags": []
     }
    }
   },
   {
    "name": "lanhu_say_edit",
    "description": "Edit message\n\nUSE THIS WHEN user says: 编辑留言, 修改消息, 更新内容\n\nPurpose: Edit published message, will record editor and edit time\n\nReturns:\n    Updated message details",
    "inputSchema": {
     "additionalProperties": false,
     "properties": {
      "url": {
       "description": "蓝湖URL（含tid和pid）",
       "type": "string"
      },
      "message_id": {
       "description": "要编辑的消息ID"
      },
      "summary": {
       "anyOf": [
        {
         "type": "string"
        },
        {
         "type": "null"
        }
       ],
       "default": null,
       "description": "新标题（可选，不传则不修改）"
      },
      "content": {
       "anyOf": [
        {
         "type": "string"
        },
        {
         "type": "null"
        }
       ],
       "default": null,
       "description": "新内容（可选，不传则不修改）"
      },
      "mentions": {
       "anyOf": [
        {
         "items": {
          "type": "string"
         },
         "type": "array"
        },
        {
         "type": "null"
        }
       ],
       "default": null,
       "description": "新@列表（可选，不传则不修改）"
      }
     },
     "required": [
      "url",
      "message_id"
     ],
     "type": "object"
    },
    "outputSchema": {
     "additionalProperties": true,
     "type": "object"
    },
    "_meta": {
     "fastmcp": {
      "tags": []
     }
    }
   },
   {
    "name": "lanhu_say_delete",
    "description": "Delete message\n\nUSE THIS WHEN user says: 删除留言, 删除消息, 移除\n\nPurpose: Delete published message\n\nReturns:\n    Delete result",
    "inputSchema": {
     "additionalProperties": false,
     "properties": {
      "url": {
       "description": "蓝湖URL（含tid和pid）",
       "type": "string"
      },
      "message_id": {
       "description": "要删除的消息ID"
      }
     },
     "required": [
      "url",
      "message_id"
     ],
     "type": "object"
    },
    "outputSchema": {
     "additionalProperties": true,
     "type": "object"
    },
    "_meta": {
     "fastmcp": {
      "tags": []
     }
    }
   },
   {
    "name": "lanhu_get_members",
    "description": "Get project collaborators list\n\nUSE THIS WHEN user says: 谁参与了, 协作者, 团队成员, 有哪些人\n\nPurpose: Get list of team members who have used Lanhu MCP tools to access this project\n\nReturns:\n    Collaborator list with first and last access time",
    "inputSchema": {
     "additionalProperties": false,
     "properties": {
      "url": {
       "description": "蓝湖URL（含tid和pid）",
       "type": "string"
      }
     },
     "required": [
      "url"
     ],
     "type": "object"
    },
    "outputSchema": {
     "additionalProperties": true,
     "type": "object"
    },
    "_meta": {
     "fastmcp": {
      "tags": []
     }
    }
   }
  ]
 }
}