
Edition Api
io.github.hiroshic9-png由 GitHub 帳號 hiroshic9-png 發布v0.3.1更新於 Oct 24, 2026
Japan business regulations, compliance, travel, protocols, memory — 10 knowledge domains, 23 tools.
概覽
本機 MCP 伺服器,為助理提供結構化的日本知識:法規、商務禮儀、旅遊、日常生活,以及橫跨 20 個領域的日語感知記憶。
- 可以做什麼
- 查詢 10 個產業的法規與旅客合規要求,並為已執行的法規檢查記錄稽核日誌、產生治理報告。檢索日本的商務禮儀、行事曆、地區差異、組織結構與外資進入指南。搜尋生活與文化主題:旅遊、飲食娛樂、日常生活、語言、災害安全、醫療、教育、房地產、稅務、銀行與簽證。儲存、召回並擷取持久記憶的片段、事實與脈絡,含日語敬語與階層分析。用一次跨領域搜尋同時涵蓋全部 20 個領域。
- 適用情境
- 適合助理需要回答在日本經營或生活的問題時使用,例如公司設立、簽證、產業監管、商務禮節或日常事務,並希望答案基於結構化條目而非通用模型知識。也適合需要日語感知持久記憶或合規稽核紀錄的代理。
- 執行需求
- 透過 npx(Node.js)以 stdio 程序在本機執行,也可經由託管 API 存取。需要能連線 EDITION API 基礎位址(EDITION_API_URL)的網路,並在 EDITION_API_KEY 中提供 API 金鑰;預設提供免費測試金鑰。免費層僅含 8 個核心領域、每天 100 次請求;全部 20 個領域需付費 Pro 方案。
安裝
在 SourceWeft 中
- 開啟 儀表板中的 Edition Api,將其新增到工作區。
- 為需要使用其工具的對話啟用該服務。
Desktop only,透過 STDIO。 STDIO 服務會啟動本機處理程序,因此需要 SourceWeft 桌面主機。
其他 MCP 客戶端
參照 儲存庫 中的啟動說明。
README
EDITION Intelligence Platform
[Glama Score] [npm] [License: MIT]
Japan Knowledge Gateway for autonomous AI agents.
20 knowledge domains · 43 MCP tools · 7 Skills Packs · 100+ REST endpoints Verified ground truth for operating in the Japanese market. Quality score: 97.8/100.
Production API: api.edition.sh — Free beta, no registration required.
Interactive Demo: Tourist Agent Demo — Watch 6 domains work together in real-time.
Why EDITION?
AI agents working in Japan hit walls that generic LLMs can't solve:
EDITION provides verified, structured intelligence that agents can use as ground truth instead of hallucinating cultural nuances, legal requirements, and procedural details.
20 Knowledge Domains
Skills Packs (MCP Skills Primitive)
Pre-built knowledge packs that bundle domain expertise, SOPs, and tool orchestration instructions. Skills-aware agents can load these to gain structured Japan operational knowledge.
→ Full Skills Pack documentation
Governance Intelligence
EDITION includes built-in compliance audit trail tools for agents operating under governance requirements. Available in both Progressive Discovery and Legacy modes.
Designed for EU AI Act (Aug 2026) and upcoming Japan AI governance compliance.
Quick Start
Option 1: npx (Recommended for Claude Desktop / Cursor)
Add to claude_desktop_config.json:
Option 2: Smithery
Option 3: REST API (Direct)
Memory API — Japanese-aware Persistent Memory
Store episodes, auto-extract structured facts with keigo analysis, social hierarchy detection, and confidence scoring.
Three-layer architecture:
- Episodes — raw conversation logs
- Facts — structured knowledge (subject→predicate→object triples)
- Context — aggregated session summaries per entity/topic
MCP Capabilities
API Endpoints
Memory
Regulation & Compliance
Business Intelligence
Market Entry
Lifestyle & Culture
Cross-Domain
Data Quality
All knowledge entries include:
source_url— Link to authoritative source (government websites, official organizations)last_verified— Date of last verificationconfidence— Verification status (verified/estimated)version— Entry version with changelog
Sources include: 厚生労働省 (MHLW), 国税庁 (NTA), 法務省 (MOJ), 国土交通省 (MLIT), 経済産業省 (METI), 日本経団連 (Keidanren), JETRO, 入管庁 (ISA).
Tech Stack
Agent Discovery
Agent Harness Compatible
EDITION is designed as a Tool Registry / Japan Knowledge Layer for agent harnesses. All 43 tools include MCP annotations (readOnlyHint, destructiveHint, idempotentHint), usage guidelines, and behavioral transparency metadata — enabling harness frameworks to auto-discover, evaluate, and integrate EDITION tools without manual configuration.
Freemium Access
Anonymous access defaults to Free tier. Register for an API key at POST /api/v1/auth/register.
Registries
Why Not Mem0 / Letta / Zep?
Those are excellent general-purpose memory tools. But they don't:
- Parse Japanese keigo levels (丁寧語 / 尊敬語 / 謙譲語)
- Detect implicit social hierarchy from honorific patterns
- Score confidence based on Japanese speech patterns (断定 vs 推測 vs 伝聞)
- Include a Japanese regulatory database with 20 knowledge domains
Japanese business context is structurally different. Agents need purpose-built infrastructure to navigate it.
License
MIT
來源:README.md,提交 74b8cc9
工具
0版本歷史
1- v0.3.1最新Sep 16, 2026
評分
已滿足適用檢查項的 50%
該評分衡量此收錄條目的完整程度,依據註冊中心條目、端點與公開來源計算得出。它不是對伺服器本身的評價,也不表示伺服器安全或目前可用:可連通性是帶日期的檢查結果,評分也不包含安裝量或使用者評價。
- 近 30 天內可連通25 分
請保持端點能夠回應 MCP 握手;SourceWeft 每 30 天探測一次。
- 已宣告認證方式10 分
請在註冊中心條目中宣告憑證請求標頭,或保持端點能夠回應探測,以便判定其認證方式。
- 已列出工具15 分
請讓伺服器在無憑證時回應 tools/list,或發布可在沙箱中無憑證啟動的軟體包。
- 描述必需15 分
- 圖示10 分
- 文件15 分
- 授權條款10 分