HallucC — Hallucination & Safety Guardian

by fredyee · remote · no sign-in to list tools · v0.1.10 · published by GitHub user fredyee

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

Changes · 5

Compare Registry entry

5 tools: 5 change things.

Scanned Oct 21, 2026

Deletes or cancels0Claude Code: deny
Can't be undone0Claude Code: ask
Changes things5Claude Code: ask: see these tools
Only reads0Claude Code: allow

1 check matched something in the tool definitions. No finding is not a safety guarantee. How the checks work.

claude mcp add --transport http hallucc https://aihcc.cloud/mcp --header 'Authorization: <your-authorization>'

Findings

Matches of fixed checks, with the text they matched.

1 warn

Warncheck_cua_code_audit
The server marks "check_cua_code_audit" read-only (readOnlyHint: true), and its own name, description or input give the class "write".
Matched text
annotations: {"readOnlyHint":true,"destructiveHint":false,"idempotentHint":true,"openWorldHint":false}; The input takes a statement or command to run (input.code), which can change anything it can reach.
What this check looks for

Tools

Open a tool for its reasons and full description. Descriptions are the publisher's own words, shown as they are.

5 tools

Changes things 5Creates or changes something that can be changed back.Claude Code: ask

check_cua_actionsComputer-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…No findings
Why it is in this group
- Nothing in its name or description says what it does, so it is counted as a write.
- The server marks it read-only (readOnlyHint: true), but an annotation never lowers a class.
Description, as the server wrote it
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。
Annotations the server declares
- readOnlyHint: true
- destructiveHint: false
- idempotentHint: false
- openWorldHint: false
check_cua_code_auditCUA 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。1 finding
Why it is in this group
- The input takes a statement or command to run (input.code), which can change anything it can reach.
- The server marks it read-only (readOnlyHint: true), but an annotation never lowers a class.
Findings
- Warn
  The server marks "check_cua_code_audit" read-only (readOnlyHint: true), and its own name, description or input give the class "write".
  Matched text
  annotations: {"readOnlyHint":true,"destructiveHint":false,"idempotentHint":true,"openWorldHint":false}; The input takes a statement or command to run (input.code), which can change anything it can reach.
  What this check looks for
Description, as the server wrote it
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。
Annotations the server declares
- readOnlyHint: true
- destructiveHint: false
- idempotentHint: true
- openWorldHint: false
check_safetyAI 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…No findings
Why it is in this group
- Nothing in its name or description says what it does, so it is counted as a write.
- The server marks it read-only (readOnlyHint: true), but an annotation never lowers a class.
Description, as the server wrote it
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）。
Annotations the server declares
- readOnlyHint: true
- destructiveHint: false
- idempotentHint: false
- openWorldHint: false
verify_agentAgent 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。No findings
Why it is in this group
- Nothing in its name or description says what it does, so it is counted as a write.
- The server marks it read-only (readOnlyHint: true), but an annotation never lowers a class.
Description, as the server wrote it
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。
Annotations the server declares
- readOnlyHint: true
- destructiveHint: false
- idempotentHint: false
- openWorldHint: true
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）。No findings
Why it is in this group
- Nothing in its name or description says what it does, so it is counted as a write.
- The server marks it read-only (readOnlyHint: true), but an annotation never lowers a class.
Description, as the server wrote it
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）。
Annotations the server declares
- readOnlyHint: true
- destructiveHint: false
- idempotentHint: false
- openWorldHint: true

This is not a safety rating, and no finding is not a safety guarantee. The groups and findings come from fixed checks of the tool definitions the server returned on the date shown, not from its code or what it does when called. How the checks work.

Install

Pinned to version 0.1.10. Values in angle brackets are yours to fill in.

claude mcp add --transport http hallucc https://aihcc.cloud/mcp --header 'Authorization: <your-authorization>'

Run in a terminal.

Claude Code permissions

Allow the reads, ask before a change, deny the deletes. Paste into .claude/settings.json.

.claude/settings.json

{
  "permissions": {
    "allow": [],
    "ask": [
      "mcp__hallucc__verify_text",
      "mcp__hallucc__verify_agent",
      "mcp__hallucc__check_cua_actions",
      "mcp__hallucc__check_cua_code_audit",
      "mcp__hallucc__check_safety"
    ],
    "deny": []
  }
}

Tools that only read are allowed, tools that change things ask first, tools that delete or cancel are denied. It assumes the server is installed as hallucc. Review it before you use it.

Keep track

Get told when this server's tools change: a new tool, a new description, a new class.

https://anlyon.dev/servers/io.github.fredyee/hallucc/changes.xml

An Atom feed. Works in any feed reader or Slack's RSS app.

Anlyon alternative

5 of these tools change something. Run those as governed actions.

Declare what an action changes, and Anlyon puts it under a shared limit, binds every approval to the exact request, and grades the outcome. It works on any HTTPS API you can describe as an action.

Anlyon does not run, proxy or govern this MCP server. A tool your own code calls directly is not routed through Anlyon and is not governed by it.

Import an OpenAPI spec →

Compare with other servers for the same product.

Source: MCP Registry · repository. Synced Oct 21, 2026, 18:34 UTC.

For agents: Markdown · JSON · report this listing