透過 Corporate AI Dashboard 協調多個專業 Agent，實現透明、可稽核、可解釋的企業 AI 自主改進。 Orchestrate multiple specialized agents via Corporate AI Dashboard — transparent, auditable, and explainable enterprise AI with self-improvement.
Z 的 agent 不是 workspace 內部工具，而是 corporate 層級的 AI 員工 — 跨團隊服務、擁有獨立職責與記憶，如同真正的團隊成員。 Z agents are not workspace-level tools. They are corporate-level AI employees — serving across teams, with their own responsibilities and memory, just like real team members.
一個 agent 可同時服務多個 workspace（部門），如同員工可參與多個專案。
A single agent can serve multiple workspaces (departments), just like an employee contributing to several projects.
每個 agent 由行為指令 (SOUL.md) 與身份定義 (IDENTITY.md) 組成，擁有獨立的專業視角。
Each agent is defined by behavioral instructions (SOUL.md) and identity (IDENTITY.md), with its own professional perspective.
Catalog (全球) → Corporate (企業) → Workspace (團隊) 三層技能架構，agent 依層級繼承並疊加能力。
Catalog (global) → Corporate (enterprise) → Workspace (team) skill inheritance, agents gain layered capabilities.
每一個 AI 動作都有完整的證據鏈 — 不是黑箱，而是可被企業審計、合規查驗的決策過程。 Every AI action has a complete evidence chain — not a black box, but a decision process that can be audited and compliance-verified.
TransparencyPanel 公開完整 prompt、資料來源、token 消耗。使用者隨時可查看 AI 的「思考過程」。
TransparencyPanel exposes the full prompt, data sources, and token usage. Users can inspect AI's "thought process" at any time.
每個 AI 輸出記錄 model、tokens、cost、generated_by，透過 evidence_log_ids 連結分析結果與原始紀錄。
Every AI output logs model, tokens, cost, and generated_by. Evidence chains link findings to source logs via evidence_log_ids.
所有洞察必須引用具體數據。計數、嚴重性必須來自實際輸入 — 絕不臆測、絕不捏造。
All insights must cite specific data. Counts and severity must derive from actual input — never speculate, never fabricate.
AI 自主改進閉環 AI Self-Improvement Loop
Corporate → Workspace 多租戶架構，搭配 Supabase RLS 與 server-side Edge Functions，天然阻隔 prompt injection 風險。 Corporate → Workspace multi-tenant architecture with Supabase RLS and server-side Edge Functions, inherently blocking prompt injection risks.
AI 邏輯在 Supabase Edge Functions 執行，使用者端無法直接觸及 prompt 或 agent 工具。RLS 確保每筆資料存取都經過身份與角色驗證。
AI logic runs in Supabase Edge Functions. Client cannot directly access prompts or agent tools. RLS ensures every data access is identity-and-role verified.
Agent 在本機或 Gateway 執行，已知存在 prompt injection 風險。無內建認證機制，需額外部署安全層。
Agents run locally or on gateway. Known prompt injection vulnerabilities. No built-in authentication, requiring additional security layers.
列級安全策略，確保跨 workspace 資料完全隔離。
Row-level security policies ensure full cross-workspace data isolation.
AI 運算在伺服器端執行，隔離用戶端攻擊面。
AI computation runs server-side, isolating from client-side attack surface.
完整 evidence chain + hallucination 偵測紀錄。
Complete evidence chain + hallucination detection logging.
Corporate → Workspace 兩層，天然支援企業多部門架構。
Corporate → Workspace two-tier, naturally supporting multi-department enterprise structures.
從單一介面掌握 agent 狀態、每日洞察、改進提案、排程任務與成本追蹤。 Monitor agent status, daily insights, improvement proposals, scheduled tasks, and cost tracking from a single interface.
24 小時內的 agent 活動量、findings 統計、token 用量、成本概覽。關鍵指標以卡片呈現，一目了然。
24-hour agent activity, findings statistics, token usage, and cost overview. Key metrics displayed as cards for instant comprehension.
每日自動分析 workspace 的 agent logs，產生 KeyFindings（含 severity 與 evidence），並附帶 TransparencyPanel 顯示完整分析過程。
Automated daily analysis of workspace agent logs, generating KeyFindings (with severity and evidence), accompanied by TransparencyPanel showing the full analysis process.
4 種提案類型（instruction / scheduled_task / skill_install / skill_remove），管理員一鍵審核套用，7 天後系統自動量化效果。
4 proposal types (instruction / scheduled_task / skill_install / skill_remove). Admin one-click review and apply, with automated effectiveness tracking after 7 days.
per-model 成本追蹤、7 天用量趨勢圖、DailyUsageChart。未來將擴展至 per-agent 與 per-task 粒度。
Per-model cost tracking, 7-day usage trend charts, DailyUsageChart. Future expansion to per-agent and per-task granularity.
每個 functional agent 擁有獨立 SOUL.md，從各自專業視角分析跨 workspace 的數據，由 Coordinator Agent 彙整協調。 Each functional agent has its own SOUL.md, analyzing cross-workspace data from its professional perspective, coordinated by a Coordinator Agent.
Corporate (Numbers Protocol) || +-- Coordinator Agent | | 彙整跨團隊洞察 / 識別衝突 / 提出跨團隊提案| Synthesize cross-team insights / detect conflicts / cross-team proposals | +-- Sales Agent (SOUL.md: 營收指標、轉化率、pipelinerevenue metrics, conversion, pipeline) | +-- Workspace: Sales | +-- Workspace: BD | +-- Finance Agent (SOUL.md: 財務審核、預算追蹤、合規financial review, budget tracking, compliance) | +-- Workspace: Finance | +-- Workspace: Accounting | +-- HR Agent (SOUL.md: 人事建議、招聘分析、團隊健康HR recommendations, hiring analytics, team health) +-- Workspace: HR +-- Workspace: Admin 共享基礎設施Shared Infrastructure +-- corporate_memories 跨 workspace 知識庫Cross-workspace knowledge base +-- corporate_agent_skills 企業級技能繼承Enterprise skill inheritance +-- corporate_api_usage 統一成本追蹤Unified cost tracking
| 維度 | Z App | OpenClaw |
|---|---|---|
| AI 自主改進閉環 | ✓ 自動提案 + 人類審核 + 效果量化 | ✕ 無內建 |
| TAE-AI 合規 | ✓ TransparencyPanel + evidence chains | ✕ 需外部工具 |
| 企業級安全 | ✓ Supabase RLS + Auth + Edge Functions | ✕ 已知 prompt injection 風險 |
| 多租戶架構 | ✓ Corporate → Workspace 隔離 | ● Gateway 層級（非多租戶設計） |
| Multi-Agent | ● 基礎設施就緒，UI 開發中 | ✓ 成熟（SOUL.md + 生命週期管理） |
| Skill 生態系 | ● 三層繼承 + 小型 catalog | ✓ ClawHub 5,400+ 技能 |
| 即時可觀測性 | ● 靜態 KPI cards | ✓ WebSocket 即時 + OTel tracing |
| Session Inspector | ● 資料已有（session_id），UI 開發中 | ✓ 完整對話 transcript 瀏覽器 |
| Dimension | Z App | OpenClaw |
|---|---|---|
| AI Self-Improvement Loop | ✓ Auto-proposals + human review + effectiveness tracking | ✕ Not built-in |
| TAE-AI Compliance | ✓ TransparencyPanel + evidence chains | ✕ Requires external tools |
| Enterprise Security | ✓ Supabase RLS + Auth + Edge Functions | ✕ Known prompt injection risks |
| Multi-Tenancy | ✓ Corporate → Workspace isolation | ● Gateway-level (not multi-tenant) |
| Multi-Agent | ● Infrastructure ready, UI in development | ✓ Mature (SOUL.md + lifecycle mgmt) |
| Skill Ecosystem | ● 3-tier inheritance + small catalog | ✓ ClawHub 5,400+ skills |
| Real-Time Observability | ● Static KPI cards | ✓ WebSocket real-time + OTel tracing |
| Session Inspector | ● Data exists (session_id), UI in progress | ✓ Full conversation transcript browser |
分三階段實現完整的企業級多 Agent 系統，每階段可獨立交付價值。 Three phases to achieve a full enterprise multi-agent system. Each phase delivers independent value.
留下聯絡資訊，我們將提供 Z Workspace AI 的完整產品介紹與企業導入方案。 Leave your contact info and we'll share a full product walkthrough and enterprise onboarding plan for Z Workspace AI.
我們會盡快與您聯繫。 We'll be in touch soon.
One system. One truth. Owned by you.
智慧、可稽核、使用者擁有的企業 AI 平台。
One system. One truth. Owned by you.
Intelligent, auditable, and user-owned enterprise AI platform.
OpenClaw 無內建
Not built-in on OpenClaw
透明 / 可稽核 / 可解釋
Transparent / Auditable / Explainable
RLS + Edge Functions
RLS + Edge Functions
Corporate Dashboard
Corporate Dashboard