WayToClawEarn
Medium impactAnthropic + Hacker News

Anthropic open source Claude for Legal: Install the legal AI plug-in in 60 seconds

Anthropic open source releases Claude for Legal, a legal AI plug-in system based on Claude Cowork and Claude Code. Covering 10+ legal areas such as commercial contract review, NDA classification, intellectual property, labor law, data privacy, etc., installation only takes 60 seconds, and is completely open source (Apache 2.0).

WayToClawEarn EditorialPublished May 15, 2026Updated Aug 8, 2026

Editorial review of public sources · AI-assisted drafting. How we work · Original source

Core conclusion

On May 14, 2026, Anthropic open sourced Claude for Legal - an AI plug-in system for legal work scenarios on Hacker News. It supports installation in two ways: Claude Cowork plug-in or Claude Code plug-in, covering 10+ legal fields including commercial contracts, corporate legal affairs, data privacy, AI governance, intellectual property, labor law, litigation, legal learning**, etc. The installation only takes 60 seconds.

Key Points

  • Published: 2026-05-14 (HN 77 points)
  • Release Format: GitHub Open Source (Apache 2.0 License)
  • Available methods: Claude Cowork plugin / Claude Code plugin / Claude Managed Agents API
  • Core Design: Cold Start Interview + CLAUDE.md Practice Profile + MCP Connector + Naming Agent
  • Highlights: Open source, neutral design, lawyer review, accurate citation and traceability

Background: The way AI enters the legal industry is changing

The legal industry has always been a "high threshold" field for AI applications - regulatory complexity, privacy protection, attorney-client privilege and other issues make it difficult to directly apply general AI tools. Although there were professional tools such as LexisNexis and CoCounsel (Thomson Reuters) on the market previously, they were usually closed SaaS products that lawyers could not customize or extend.

Anthropic has adopted a completely different strategy for this open source action: instead of selling a closed "legal AI product", it is open source a set of "legal AI recipes"** - system prompts, skill plug-ins, and MCP connectors are all publicly available, and lawyers and law firm IT teams can freely customize them.

Core function disassembly

Functional dimensionSpecific contentWho is useful
Commercial Contract Review AgentVendor Agreement Reviewer: Automatically review MSA against your playbookCompany Legal Team
NDA automatic classificationGREEN/YELLOW/RED three-level classification, leaving only the tough stuff to the lawyersAny business dealing with a large number of NDAs
Data Privacy DSAR ProcessingAutomated responses to data subject access requestsPrivacy Compliance Team
AI Governance ReviewAutomatic scanning for AI product complianceLegal affairs for AI startups
M&A Due DiligenceTabular data room review, each note accurately cites the sourceLaw Firm M&A Group
Law School Study ToolsMock classroom Q&A, Bar exam practice, pre-test predictionsLaw School Students
Intellectual Property AnalysisClaim chart construction, IP portfolio analysisIntellectual Property Team

Technical architecture highlights

1. Cold start interview + practice profile

Each time you install the plugin, conduct a cold start interview to understand how the attorney/law firm works—contract template preferences, signature thresholds, common clause handling patterns, etc. This information is written into the CLAUDE.md practice file, and all subsequent Agents share the same work context.

2. MCP connector system

The plug-in has multiple built-in MCP (Model Context Protocol) connectors to directly connect to lawyers’ daily tools:

  • Slack — contract expiration reminder, approval request
  • Google Drive — automatic retrieval of contract documents
  • Box/iManage — Legal document management system
  • Ironclad/DocuSign — Contract Lifecycle Management
  • Everlaw — eDiscovery Platform
  • CourtListener — case database

Claude for Legal MCP

3. Naming Agent system

Each Agent has an independent name and a command entry, which can be directly called in chat using /agent-name. For example, /commercial-legal:review initiates a supplier agreement review, /corporate-legal:tabular-review initiates a data room due diligence form.

Security design: not replacing lawyers, but accelerating lawyers

Claude for Legal is designed with two key security features:

  1. Every output is a draft - All analyzes and suggestions generated by Agent will be marked as "draft before lawyer review" and are not legal advice
  2. Accurate citation tracing - Every modification suggestion in the contract review and every conclusion in the due diligence form will be accompanied by the precise citation location of the original document.

This is in sharp contrast to the "black box output" approach of many AI tools. For the legal industry, advice that cannot be traced is equivalent to no advice.

Implications for content creators

Although Claude for Legal focuses on legal scenarios, it demonstrates an important product model: Using open source + Agent plug-ins to allow AI to penetrate into professional vertical fields. This model also has direct implications for content creators and automated workflow builders:

  • Vertical Agent plug-in mode is replacing the general large model dialogue and becoming the standard form for AI implementation
  • Claude Cowork / Claude Code is an Agent runtime platform that supports the free combination of skill plug-ins in any field.
  • Open source license (Apache 2.0) means you can Fork this project and change the legal plug-in to "Audit Plug-in", "Compliance Plug-in", "Content Audit Plug-in"

Want to learn how to use Claude Code to build your own vertical Agent plug-in? Look for inspiration from existing cases and experiences.

Internal link guidance

View source →

Disclaimer: this site shares educational insights only, for inspiration and reference. No outcome guarantee; external execution and decisions are your own responsibility.