Configs2026年7月18日·1 分钟阅读

Claude for Legal — AI-Powered Legal Workflow Plugins by Anthropic

An official suite of plugins from Anthropic designed for legal professionals. Covers contract analysis, case research, document drafting, and compliance review with domain-specific safety guardrails.

Agent 就绪

先审查再安装

这个资产需要先审查。复制的指令会要求 Agent dry-run、列出写入项,确认后再继续。

Needs Confirmation · 66/100策略:需确认
Agent 入口
任意 MCP/CLI Agent
类型
Skill
安装
Single
信任
信任等级:Established
入口
Claude for Legal Guide
先审查命令
npx -y tokrepo@latest install 882e526a-82c6-11f1-9bc6-00163e2b0d79 --target codex

先 dry-run,确认写入项后再运行此命令。

Introduction

Claude for Legal is an official plugin suite from Anthropic providing specialized AI capabilities for legal professionals within Claude Code. It covers contract analysis, case research, document drafting, and compliance review, each with domain-specific safety guardrails that flag when human review is required.

What Claude for Legal Does

  • Analyzes contracts to identify key terms, obligations, deadlines, and non-standard clauses
  • Searches and summarizes case law, statutes, and regulatory guidance for legal questions
  • Drafts memoranda, briefs, and correspondence using jurisdiction-aware templates
  • Reviews policies against regulatory frameworks to identify compliance gaps
  • Flags legal uncertainty and marks output requiring attorney review

Architecture Overview

Four modular plugins share a common legal reasoning layer. The contract plugin segments documents into clauses and evaluates each against standard and non-standard patterns. The research plugin queries public legal databases by jurisdiction and recency. All plugins route through a safety layer that appends confidence indicators and review flags where certainty falls below a configurable threshold.

Self-Hosting & Configuration

  • Install via claude install anthropics/claude-for-legal
  • Configure target jurisdictions in .claude-legal.yml to scope research and drafting
  • Set CLAUDE_LEGAL_REVIEW_THRESHOLD to control when mandatory review flags appear
  • Citation format defaults to Bluebook; OSCOLA, AGLC, McGill selectable per session
  • Supports document management integration via configurable input/output paths

Key Features

  • Domain-specific safety guardrails that flag uncertain conclusions for human review
  • Jurisdiction-aware drafting with correct citation format, terminology, and statutory references
  • Clause-level contract analysis with per-provision risk scoring
  • Audit trail generation logging each analysis step for compliance purposes
  • Structured case research output organized by relevance, jurisdiction, and date

Comparison with Similar Tools

  • Harvey AI — full-platform legal AI; Claude for Legal integrates into the Claude Code workflow as plugins
  • CoCounsel (Thomson Reuters) — tied to Westlaw; Claude for Legal works with public databases and local files
  • Lexis+ AI — embedded in LexisNexis; Claude for Legal is vendor-neutral on user-provided documents
  • Spellbook (Rally) — contract review only; Claude for Legal adds research, compliance, and drafting

FAQ

Q: Does Claude for Legal provide legal advice? A: No. It surfaces information and drafts documents for attorney review. Output includes disclaimers and review flags where appropriate.

Q: Which jurisdictions are supported? A: US federal and state, UK, EU, and Canadian law. Additional jurisdictions can be added via statutory references in the config file.

Q: How does it handle confidential documents? A: Documents are processed via the Claude API with Anthropic's data handling policies. Nothing is stored beyond the session unless you configure local caching.

Sources

讨论

登录后参与讨论。
还没有评论,来写第一条吧。

相关资产