Law MCP Server Specification

This repository will host an MCP server that uses 法令API Version 2 (e-Gov) to fetch statute data and help check the consistency between internal documents and the referenced laws. The document below captures the target capabilities and operational expectations before implementation.

Overview

• Provide MCP tools that surface law data from the official API and perform document-to-law consistency checks.
• Enable knowledge workers to verify whether policy drafts, contracts, or memos align with authoritative legal text.
• Favor transparent outputs that include sources (LawID, article numbers, URLs) and the reasoning steps used during checks.

External Data Source

• Base: https://laws.e-gov.go.jp/api/2/
• Common endpoints (see swagger for full schema):
  • GET /law_data/{law_id_or_num_or_revision_id} – fetch law structure and articles.
  • GET /keyword?keyword={keyword} – search laws by keyword.
  • GET /laws?law_title={title} – search laws by title.
• Response format: JSON (includes meta, LawName, Articles, etc.). Respect official rate limits; treat 429/503 as retryable with backoff.

MCP Capabilities

• search_laws – Input: keyword (string). Output: list of LawID, title, and promulgation date.
• fetch_law – Input: lawId (string), optional revisionDate. Output: normalized law JSON.
• check_consistency – Input: documentText, lawIds (required). Output: matched citations, conflicting passages, and similarity scores.
• summarize_law – Input: lawId, optional articles list. Output: concise article summary with paragraph text.

Consistency Check Workflow

• Normalize the incoming document (segment by sentence/section, detect cited articles like “第○条”).
• Resolve target laws: use lawIds provided or run search_laws to suggest candidates.
• Fetch required law texts via fetch_law; cache responses per LawID to reduce API load.
• Align document segments to law articles using string similarity and citation hints; note exact article numbers when present.
• Produce findings: for each segment, mark status (aligned, potential_mismatch, not_found), include article references, and show snippets from both sides.
• Provide remediation suggestions (e.g., cite correct article, adjust wording) without altering the source document automatically.

Server Behavior & Error Handling

• Map API errors to MCP-friendly errors with actionable messages (e.g., missing LawID, upstream 429, malformed parameters).
• Use exponential backoff on 429/503 and surface retry-after hints when present.
• Validate inputs early: reject empty documentText, overly long queries, or unsupported lawId formats with clear guidance.
• Log tool calls and upstream URLs for debugging; avoid storing document contents longer than the session.

Configuration

• Environment variables (to be wired during implementation):
  • LAW_API_BASE (default https://laws.e-gov.go.jp/api/2/)
  • HTTP_TIMEOUT_MS (default 15000)
  • CACHE_TTL_SECONDS (default 900)
• Add .env.example later with the above keys; do not commit secrets.

Implementation Notes

• Suggested stack: Node.js with a lightweight stdio JSON-RPC bridge for MCP compatibility, undici/node-fetch for HTTP, and a lightweight in-memory cache (Map/LRU). TypeScript preferred for schema safety using the Swagger spec.
• Define TypeScript types for API responses (LawData, Article, SearchResult) to enforce strict parsing.
• Keep business logic pure and testable (e.g., citation extraction, alignment scoring) independent of I/O.
• Expose a health endpoint or MCP tool (e.g., ping) for quick readiness checks.

Getting Started

• Requirements: Node.js 18+.
• Install dependencies: npm install.
• Build: npm run build.
• Copy .env.example to .env and adjust if needed.
• Run server over stdio (JSON-RPC): npm start (or npm run dev for ts-node).
• Configure via environment variables in .env (see Configuration section). The server registers tools search_laws, fetch_law, check_consistency, and summarize_law.
• Quality: npm run lint (ESLint) / npm run format (Prettier).

Claude Desktop configuration

• Install globally: npm install -g law-mcp-server.
• Add to your claude_desktop_config.json:

{
  "mcpServers": {
    "law-mcp-server": {
      "command": "law-mcp-server"
    }
  }
}

If you are installing from a local clone instead of the published package, run npm install && npm run build and then npm link so the law-mcp-server command is available on your PATH for Claude Desktop.

Usage Examples (conceptual)

• Search and fetch: “Search for 個人情報保護 and show the latest articles.” → calls search_laws then fetch_law.
• Consistency check: "Check this draft against 労働基準法 Articles 24 and 37; highlight mismatches." → calls search_laws to get LawID, then check_consistency with lawIds=[...].

Skills

This repository includes domain-specific skills that demonstrate effective usage patterns for law-mcp-server tools. Skills provide comprehensive guides on how to leverage the server's capabilities for specific use cases.

Available Skills

Digital Marketing Law Skill (skills/digital-marketing-law/)

A comprehensive guide for using law-mcp-server to reference and verify compliance with laws related to digital marketing activities in Japan. This skill covers:

• Display Regulations: Misleading Representation Prevention Act (景品表示法), Specified Commercial Transactions Act (特商法), Consumer Contract Act
• Personal Information & Tracking: Personal Information Protection Act (個人情報保護法), Telecommunications Business Act (電気通信事業法), Specified Electronic Mail Act
• Platform Regulations: Digital Platform Transparency Act, Provider Liability Act
• Industry-Specific Laws: Pharmaceutical Affairs Act (薬機法), Financial Instruments and Exchange Act (金商法)
• Intellectual Property: Copyright Act, Trademark Act, Unfair Competition Prevention Act
• Competition Law: Antimonopoly Act (独占禁止法)

Key Features:

• Search patterns for formal names, abbreviations, and article numbers
• 5 practical workflows (privacy policy creation, ad review, email marketing, platform transactions, amendment tracking)
• Real-world use cases for JIAA/APTI activities, client proposals, and compliance checks
• Common Q&A (Cookie consent, influencer marketing, comparative advertising, AI-generated content, retargeting)

Usage:

1. Read the skill file: skills/digital-marketing-law/digital-marketing-law-SKILL.md
2. Reference the appropriate workflow for your task
3. Use the provided search keywords and tool sequences
4. Follow the best practices for law searches and consistency checks

Using Skills with Claude

To enable Claude to use these skills effectively:

1. With Claude Desktop: Skills in this repository are automatically available when the law-mcp-server is configured
2. With Claude API: Include the skill content in your system prompts or as reference documentation
3. Custom Integration: Point Claude to the skills directory in your MCP server configuration

Skills enhance Claude's ability to:

• Choose the right tools for specific legal queries
• Use appropriate search keywords (formal names vs. abbreviations)
• Apply domain knowledge for effective law searches
• Structure multi-step legal compliance checks
• Provide context-aware recommendations

Validation Plan (to implement)

• Unit tests for citation parsing, article alignment scoring, and API response normalization.
• Integration tests mocking the 法令API to cover success, 404, 429/503 retry, and malformed LawID cases.
• Integration test runner: npm run test (uses undici MockAgent; no network).
• Manual smoke: run MCP client (e.g., Claude Desktop) to issue search_laws and check_consistency commands.

This specification is the starting point; refine it as implementation details solidify while keeping parity with the official 法令API Version 2 documentation.