Citation Safe Verifier

Verifies legal citations vs primary sources: existence, quote match, proposition support.

我该使用它吗

质量与安全性

A
描述质量
100%
模式完整度
99%
命名质量
95%
投毒风险
100%
权限匹配度
100%
协议合规性
100%

基于对工具定义和协议合规性的自动分析。

上下文开销

~1,549token 数(工具定义)
~843 B典型响应大小
对注意力有中等影响(占 128k 上下文窗口的 1.21%)

这是每次将服务器的工具加载到模型上下文窗口时所消耗的大致 token 数。数值越高,可用于其他任务的注意力就越少。

安装

一键安装

将以下内容添加到你的 `claude_desktop_config.json` 文件中:

{
  "mcpServers": {
    "verifier": {
      "url": "https://citationsafe.com/api/mcp"
    }
  }
}

远程端点

https://citationsafe.com/api/mcpstreamable-http

它能做什么

工具清单

工具(8)

🟢 只读🟡 写入🔴 删除⚪ 未知
⚪verify_citation(text)

Checks every legal citation in the given text against primary sources: does the cited case exist (Layer 1, CourtListener database lookup — deterministic, no LLM), does a quoted passage appear in the source opinion (Layer 2), and — on paid CitationSafe tiers — does the case support the proposition it's cited for (Layer 3). Returns VERIFIED / UNCONFIRMED / UNCONFIRMED — OUTSIDE COVERAGE / SOURCE UNAVAILABLE per citation, never a bare true/false, with source links. Not legal advice.

输入模式

{
  "type": "object",
  "properties": {
    "text": {
      "type": "string",
      "minLength": 1,
      "maxLength": 64000,
      "description": "Raw text containing one or more legal citations to verify"
    }
  },
  "required": [
    "text"
  ],
  "$schema": "http://json-schema.org/draft-07/schema#"
}
⚪verify_document(url)

Fetches a document (PDF, DOCX, or plain text) from a URL and runs the same 3-layer citation verification as verify_citation. Use this when you have a link to a brief rather than pasted text.

输入模式

{
  "type": "object",
  "properties": {
    "url": {
      "type": "string",
      "format": "uri",
      "description": "URL of the PDF, DOCX, or text document to fetch and verify"
    }
  },
  "required": [
    "url"
  ],
  "$schema": "http://json-schema.org/draft-07/schema#"
}
🟢get_verification(id)

Retrieves a previously run CitationSafe verification (by the verificationId returned from verify_citation/verify_document, or from a citationsafe.com/v/{id} share link) and returns its full per-citation results.

输入模式

{
  "type": "object",
  "properties": {
    "id": {
      "type": "string",
      "format": "uuid",
      "pattern": "^([0-9a-fA-F]{8}-[0-9a-fA-F]{4}-[1-8][0-9a-fA-F]{3}-[89abAB][0-9a-fA-F]{3}-[0-9a-fA-F]{12}|00000000-0000-0000-0000-000000000000|ffffffff-ffff-ffff-ffff-ffffffffffff)$",
      "description": "Verification ID"
    }
  },
  "required": [
    "id"
  ],
  "$schema": "http://json-schema.org/draft-07/schema#"
}
⚪batch_verify_citations(texts)

Runs the same 3-layer citation verification as verify_citation across multiple texts in a single call. Each item is billed against your CitationSafe usage exactly like an individual verify_citation call — batching never bypasses your plan's monthly limit. Max 10 items per batch, 64,000 characters per item. Requires a signed-in CitationSafe account (Authorization: Bearer <access_token>) — anonymous batch calls are not supported, use verify_citation for anonymous single-text checks.

输入模式

{
  "type": "object",
  "properties": {
    "texts": {
      "minItems": 1,
      "maxItems": 10,
      "type": "array",
      "items": {
        "type": "string",
        "minLength": 1,
        "maxLength": 64000
      },
      "description": "Array of texts to verify, each processed as a separate verification (max 10 items, 64,000 chars each)"
    }
  },
  "required": [
    "texts"
  ],
  "$schema": "http://json-schema.org/draft-07/schema#"
}
🟢search_sanctions(query, court)

Searches CitationSafe's database of real court sanctions/orders arising from AI-hallucinated legal citations, by court name, party/case name keyword, or free-text keyword. Read-only, public data — same dataset backing citationsafe.com/sanctions-database.

输入模式

{
  "type": "object",
  "properties": {
    "query": {
      "description": "Free-text keyword to match against the case name",
      "type": "string",
      "minLength": 1,
      "maxLength": 200
    },
    "court": {
      "description": "Filter by court name (partial match, e.g. \"S.D.N.Y.\")",
      "type": "string",
      "minLength": 1,
      "maxLength": 200
    }
  },
  "$schema": "http://json-schema.org/draft-07/schema#"
}
🟢is_this_case_real(citation)

Deterministic existence check (Layer 1 only, no LLM) for one legal citation against CourtListener's primary-source database. Answers the single question "is this case real" for one citation — for multi-citation quote-checked verdicts across a whole brief, use verify_brief instead. Free tier: 5 checks/day per IP, no key required. For unlimited access, pass an API key as "Authorization: Bearer <key>" (contact [email protected] to request one).

输入模式

{
  "type": "object",
  "properties": {
    "citation": {
      "type": "string",
      "minLength": 1,
      "maxLength": 2000,
      "description": "A single citation, e.g. \"410 U.S. 113\" or \"Roe v. Wade, 410 U.S. 113 (1973)\""
    }
  },
  "required": [
    "citation"
  ],
  "$schema": "http://json-schema.org/draft-07/schema#"
}
🟢verify_brief(text)

Runs existence (Layer 1) and quote-match (Layer 2) checks — both deterministic, no LLM — on every citation found in the given text and returns one verdict per citation. The multi-citation counterpart to is_this_case_real. Proposition support (Layer 3) is not included on this free/API-key surface; sign in at citationsafe.com for full 3-layer verification via verify_citation. Free tier: 5 checks/day per IP (one verify_brief call = one check, regardless of how many citations the text contains); an API key bypasses the daily cap.

输入模式

{
  "type": "object",
  "properties": {
    "text": {
      "type": "string",
      "minLength": 1,
      "maxLength": 64000,
      "description": "Raw text (brief excerpt, motion, memo) containing one or more legal citations"
    }
  },
  "required": [
    "text"
  ],
  "$schema": "http://json-schema.org/draft-07/schema#"
}
🟢search_phantoms(query)

Searches CitationSafe's phantom-citation registry — fabricated case names/citations that have shown up in real, sanctioned AI-hallucination incidents — by case name or the AI system that produced them. Read-only, public data. Returns an empty array (not an error) if nothing matches or the registry has no entries yet.

输入模式

{
  "type": "object",
  "properties": {
    "query": {
      "type": "string",
      "minLength": 1,
      "maxLength": 200,
      "description": "Free-text match against the phantom case name or the AI system that generated it"
    }
  },
  "required": [
    "query"
  ],
  "$schema": "http://json-schema.org/draft-07/schema#"
}

社区

评价此服务器

证据

最近观测

已验证未记录版本8 个工具
已验证未记录版本8 个工具