HallucC — Hallucination & Safety Guardian

Working · 2.8 s

io.github.fredyee/halluccv0.1.10

Fact-check LLM outputs with cited sources. API key: aihcc.cloud/keys.

Security

Source Website Registry entry

Maintain this server? Claim it

Try it

Call this server's tools from your browser. Nothing to install.

https://aihcc.cloud/mcp

Authorization (required)

Bearer <HallucC API key> — create at aihcc.cloud/keys after signup; free daily quota included（注册后于 /keys 页创建，免费套餐每日有额度）

Requests go through mcptry because browsers can't call most MCP servers directly. Header values are sent to the server for that request only and are never stored or logged.

Tool list from our last check. Connect live to refresh it.

check_cua_actions Read-only

Computer-Use Agent 动作风险分级。提交动作轨迹 → 逐步返回 L0(放行)/L1(放行+记录)/L2(需确认)/L3(阻断)/takeover(凭据场景挂起等用户亲手输入) 裁决、matched_rules、outcome，及 L0-L3 汇总计数；响应含 ruleset_version（规则库版本）与 disclaimer（本结果仅为分级参考，未执行实际拦截）。覆盖：敏感路径/破坏性命令/不可逆元素语义、域名分级（未知域导航 L2、黑名单域与敏感粘贴到未知域 L3、支付金额超阈值 L3）、凭据词表与支付/登录域 takeover、task_scope 任务意图偏离（越界升一级，只能收窄）、多智能体委托链字段（每条动作可选 chain_id/agent_id/parent_agent_id/delegation_depth，给了则服务端追加链级跨跳污点分析：跨 agent 传播时聚合级别抬到 L3 并置 risk_diluted，返回在 chain_trace 字段；未给则零影响）。纯规则判定，不调 LLM、不耗额度。用于在 agent 执行前/后做安全自检。

actionsarray, required

待分级的动作序列（1~500 条）

session_idstring

给了则把分级结果写入审计表，可在 /cua/audit 回放

task_scopeobject

任务意图对齐（收窄机制）：动作越出 scope 时升一级处置（L0→L1、L1→L2、L2 保持但追加原因、L3 不变）。只能升级不能降级；不传则零影响。

Try it

Call this server's tools from your browser. Nothing to install.

https://aihcc.cloud/mcp

Authorization (required)

Bearer <HallucC API key> — create at aihcc.cloud/keys after signup; free daily quota included（注册后于 /keys 页创建，免费套餐每日有额度）

Requests go through mcptry because browsers can't call most MCP servers directly. Header values are sent to the server for that request only and are never stored or logged.

Tool list from our last check. Connect live to refresh it.

check_cua_actions Read-only

actionsarray, required

待分级的动作序列（1~500 条）

session_idstring

给了则把分级结果写入审计表，可在 /cua/audit 回放

task_scopeobject

任务意图对齐（收窄机制）：动作越出 scope 时升一级处置（L0→L1、L1→L2、L2 保持但追加原因、L3 不变）。只能升级不能降级；不传则零影响。

Tools (5)

check_cua_actions

read-only

Argument	Type	Description
actions*	array	待分级的动作序列（1~500 条）
session_id	string	给了则把分级结果写入审计表，可在 /cua/audit 回放
task_scope	object	任务意图对齐（收窄机制）：动作越出 scope 时升一级处置（L0→L1、L1→L2、L2 保持但追加原因、L3 不变）。只能升级不能降级；不传则零影响。

Input schema (JSON)

{
  "type": "object",
  "$schema": "http://json-schema.org/draft-07/schema#",
  "required": [
    "actions"
  ],
  "properties": {
    "actions": {
      "type": "array",
      "items": {
        "type": "object",
        "properties": {
          "x": {
            "type": "number",
            "description": "坐标式点击 X"
          },
          "y": {
            "type": "number",
            "description": "坐标式点击 Y"
          },
          "app": {
            "type": "string",
            "description": "target_app 别名"
          },
          "url": {
            "type": "string",
            "description": "target_url 别名"
          },
          "path": {
            "type": "string",
            "description": "target_path 别名"
          },
          "text": {
            "type": "string",
            "description": "type/key/paste 的内容"
          },
          "action": {
            "type": "string",
            "description": "动作名（click/press/hotkey/write/move…），省略则按坐标推断"
          },
          "element": {
            "type": "string",
            "description": "target_element 别名"
          },
          "thought": {
            "type": "string",
            "description": "agent 推理文本（审计留痕）"
          },
          "agent_id": {
            "type": "string",
            "description": "P1-9 本动作所属 Agent 名（如 research），配合 registered_agents 做注册校验"
          },
          "chain_id": {
            "type": "string",
            "description": "P1-9 多智能体委托链 id（uuid hex）；给了则审计按 chain_id 跨会话聚合，可在 /cua/audit 看链路视图"
          },
          "coordinate": {
            "type": "array",
            "items": [
              {
                "type": "integer",
                "maximum": 9007199254740991,
                "minimum": -9007199254740991
              },
              {
                "type": "integer",
                "maximum": 9007199254740991,
                "minimum": -9007199254740991
              }
            ],
            "maxItems": 2,
            "minItems": 2,
            "description": "[x, y] 二元组（与 x/y 二选一）",
            "additionalItems": false
          },
          "target_app": {
            "type": "string",
            "description": "目标应用"
          },
          "target_url": {
            "type": "string",
            "description": "浏览器动作的目标 URL（域名分级 / 未知域导航 / 粘贴到未知域规则用）"
          },
          "task_scope": {
            "type": "object",
            "required": [
              "task"
            ],
            "properties": {
              "task": {
                "type": "string",
                "minLength": 1,
                "description": "声明的任务意图（如「整理本周会议纪要」）"
              },
              "allowed_apps": {
                "type": "array",
                "items": {
                  "type": "string"
                },
                "description": "允许的前台应用白名单"
              },
              "allowed_domains": {
                "type": "array",
                "items": {
                  "type": "string"
                },
                "description": "允许访问的域名白名单（后缀匹配含子域）"
              },
              "forbidden_elements": {
                "type": "array",
                "items": {
                  "type": "string"
                },
                "description": "禁止出现的元素标签（最高优先级）"
              }
            },
            "description": "P1-5 任务允许范围声明（服务端按动作读取；越界升一级，只能收窄）"
          },
          "target_path": {
            "type": "string",
            "description": "文件操作路径 / 终端命令路径"
          },
          "action_input": {
            "type": "string",
            "description": "ReAct 形态输入别名 → text"
          },
          "target_element": {
            "type": "string",
            "description": "目标元素标签"
          },
          "parent_agent_id": {
            "type": "string",
            "description": "P1-9 委托方 Agent 名；跨 agent 边界的传播链由此配对"
          },
          "delegation_depth": {
            "type": "integer",
            "maximum": 9007199254740991,
            "minimum": 0,
            "description": "P1-9 委托深度（根 Agent=0）；超 delegation_max_depth 升 L2"
          }
        },
        "description": "单条 CUA 动作"
      },
      "maxItems": 500,
      "minItems": 1,
      "description": "待分级的动作序列（1~500 条）"
    },
    "session_id": {
      "type": "string",
      "maxLength": 64,
      "minLength": 1,
      "description": "给了则把分级结果写入审计表，可在 /cua/audit 回放"
    },
    "task_scope": {
      "type": "object",
      "required": [
        "task"
      ],
      "properties": {
        "task": {
          "type": "string",
          "minLength": 1,
          "description": "声明的任务意图（如「整理本周会议纪要」）"
        },
        "allowed_apps": {
          "type": "array",
          "items": {
            "type": "string"
          },
          "description": "允许的前台应用白名单"
        },
        "allowed_domains": {
          "type": "array",
          "items": {
            "type": "string"
          },
          "description": "允许访问的域名白名单（后缀匹配含子域）"
        },
        "forbidden_elements": {
          "type": "array",
          "items": {
            "type": "string"
          },
          "description": "禁止出现的元素标签（最高优先级）"
        }
      },
      "description": "任务意图对齐（收窄机制）：动作越出 scope 时升一级处置（L0→L1、L1→L2、L2 保持但追加原因、L3 不变）。只能升级不能降级；不传则零影响。"
    }
  }
}

check_cua_code_audit

CUA 代码审计：静态分析 Agent 源码，检测危险导入（subprocess/os.system 等）、权限边界（eval/exec）、注入面（字符串拼接 shell）、危险默认值、沙箱缺失。纯规则扫描，不调 LLM、不耗额度。支持单文件或多文件扫描。

read-only

Argument	Type	Description
code	string	单文件模式：要扫描的源代码文本（≤50000 字符），与 files 二选一
files	array	多文件模式：[{path, content, language?}] 数组，与 code 二选一
filename	string	单文件模式：文件名，用于报告中标识（默认 input.py）
language	string	语言提示，auto 时按 filename 扩展名推断（默认 auto）

Input schema (JSON)

{
  "type": "object",
  "$schema": "http://json-schema.org/draft-07/schema#",
  "properties": {
    "code": {
      "type": "string",
      "maxLength": 50000,
      "description": "单文件模式：要扫描的源代码文本（≤50000 字符），与 files 二选一"
    },
    "files": {
      "type": "array",
      "items": {
        "type": "object",
        "required": [
          "path",
          "content"
        ],
        "properties": {
          "path": {
            "type": "string",
            "minLength": 1
          },
          "content": {
            "type": "string",
            "maxLength": 50000
          },
          "language": {
            "enum": [
              "auto",
              "python",
              "javascript",
              "typescript"
            ],
            "type": "string"
          }
        }
      },
      "description": "多文件模式：[{path, content, language?}] 数组，与 code 二选一"
    },
    "filename": {
      "type": "string",
      "description": "单文件模式：文件名，用于报告中标识（默认 input.py）"
    },
    "language": {
      "enum": [
        "auto",
        "python",
        "javascript",
        "typescript"
      ],
      "type": "string",
      "description": "语言提示，auto 时按 filename 扩展名推断（默认 auto）"
    }
  }
}

check_safety

AI 安全网关检测。对 prompt + output 跑 OWASP LLM Top 10 三层检测：40+ 特征规则引擎（注入/越狱/有害内容/敏感信息泄露/欺诈）+ 可选 LLM 深度分析 + 可选幻觉检测。返回 passed、risk_score、risk_level、threats[]（category/severity/match_context/recommendation）、quota。fast=true 走纯规则 <10ms 模式。耗 detect 额度（fast 亦计入）。

read-only

Argument	Type	Description
fast	boolean	true 走纯规则快速模式（<10ms，无 LLM）→ /guard/check-fast；默认 false 走完整 /guard/check
output*	string	AI 响应（≥1 字，≤50000 字）
prompt	string	用户原始输入（默认空，≤50000 字）
check_hallucination	boolean	是否同时检测幻觉，默认 false

Input schema (JSON)

{
  "type": "object",
  "$schema": "http://json-schema.org/draft-07/schema#",
  "required": [
    "output"
  ],
  "properties": {
    "fast": {
      "type": "boolean",
      "description": "true 走纯规则快速模式（<10ms，无 LLM）→ /guard/check-fast；默认 false 走完整 /guard/check"
    },
    "output": {
      "type": "string",
      "maxLength": 50000,
      "minLength": 1,
      "description": "AI 响应（≥1 字，≤50000 字）"
    },
    "prompt": {
      "type": "string",
      "maxLength": 50000,
      "description": "用户原始输入（默认空，≤50000 字）"
    },
    "check_hallucination": {
      "type": "boolean",
      "description": "是否同时检测幻觉，默认 false"
    }
  }
}

verify_agent

Agent 输出与轨迹自检。对最终回答跑逐声明幻觉核验，并对执行轨迹做六维结构化评估（事实性/来源/指令合规/工具声明一致/任务完成/反思）。返回 main.claims、steps[].checks、quota。耗 detect 额度。

read-only

Argument	Type	Description
task	string	原始任务指令（task_completion 裁判用）
speed	string	速度模式：fast\|standard\|deep，默认 standard
steps	array	agent 执行轨迹，逐步六维评估
domain	string	领域模式：general\|medical\|legal\|finance\|education\|government
agent_output*	string	agent 最终输出文本（≥1 字，≤50000 字）
tool_schemas	object	工具 Schema（与 trajectory 端点对齐）

Input schema (JSON)

{
  "type": "object",
  "$schema": "http://json-schema.org/draft-07/schema#",
  "required": [
    "agent_output"
  ],
  "properties": {
    "task": {
      "type": "string",
      "description": "原始任务指令（task_completion 裁判用）"
    },
    "speed": {
      "enum": [
        "fast",
        "standard",
        "deep"
      ],
      "type": "string",
      "description": "速度模式：fast|standard|deep，默认 standard"
    },
    "steps": {
      "type": "array",
      "items": {
        "type": "object",
        "properties": {
          "step": {
            "type": "integer",
            "maximum": 9007199254740991,
            "minimum": -9007199254740991,
            "description": "步序号，默认 1"
          },
          "tool": {
            "type": "string",
            "description": "该步调用的工具名"
          },
          "tool_input": {
            "type": "string",
            "description": "工具输入"
          },
          "agent_claim": {
            "type": "string",
            "description": "该步 agent 的声明"
          },
          "tool_output": {
            "type": "string",
            "description": "工具输出"
          }
        },
        "description": "agent 轨迹单步"
      },
      "description": "agent 执行轨迹，逐步六维评估"
    },
    "domain": {
      "enum": [
        "general",
        "medical",
        "legal",
        "finance",
        "education",
        "government"
      ],
      "type": "string",
      "description": "领域模式：general|medical|legal|finance|education|government"
    },
    "agent_output": {
      "type": "string",
      "maxLength": 50000,
      "minLength": 1,
      "description": "agent 最终输出文本（≥1 字，≤50000 字）"
    },
    "tool_schemas": {
      "type": "object",
      "description": "工具 Schema（与 trajectory 端点对齐）",
      "propertyNames": {
        "type": "string"
      },
      "additionalProperties": {}
    }
  }
}

verify_text

逐声明幻觉核验。输入文本 → 提取声明 → 逐声明验证（supported/refuted/unverified）+ 来源引用。返回红/黄/绿汇总、每条声明的状态/置信度/理由/来源、citations、quota。耗 detect 额度（与 Web 端同源）。

read-only

Argument	Type	Description
text*	string	待核验的文本
model	string	被测模型名（选填，分享卡片展示用）
speed	string	速度模式：fast\|standard\|deep，默认 standard
domain	string	领域模式：general\|medical\|legal\|finance\|education\|government
strict	boolean	严格模式（自我一致性），不传走全局默认

Input schema (JSON)

{
  "type": "object",
  "$schema": "http://json-schema.org/draft-07/schema#",
  "required": [
    "text"
  ],
  "properties": {
    "text": {
      "type": "string",
      "maxLength": 50000,
      "minLength": 1,
      "description": "待核验的文本"
    },
    "model": {
      "type": "string",
      "maxLength": 50,
      "description": "被测模型名（选填，分享卡片展示用）"
    },
    "speed": {
      "enum": [
        "fast",
        "standard",
        "deep"
      ],
      "type": "string",
      "description": "速度模式：fast|standard|deep，默认 standard"
    },
    "domain": {
      "enum": [
        "general",
        "medical",
        "legal",
        "finance",
        "education",
        "government"
      ],
      "type": "string",
      "description": "领域模式：general|medical|legal|finance|education|government"
    },
    "strict": {
      "type": "boolean",
      "description": "严格模式（自我一致性），不传走全局默认"
    }
  }
}