Three-Layer Architecture
Workbench / Workspace / Organization — three perspectives, boundaries, and relationships
Think of Evose as three buildings stacked on top of each other. Each layer serves a different audience and does one thing. This structure is what sets Evose apart from "single-purpose AI tools".
Task · Library · Schedule — IM-style hub for every Agent, Workflow, and model
- Task
- Library
- Schedule
- Workspace entry
Layer 1 · Workbench (end-user side)
Who it's for: any member of the organization — regular users, business users, cross-department collaborators.
Core action: use. They don't create or configure; they only call AI capabilities that have already been set up.
Four top-level menus:
| Menu | Purpose |
|---|---|
| Task | Main entry · Chat with or trigger any accessible Agent / Chatflow / Workflow / model · IM-style aggregation |
| Library | Cross-user content space · Share / like / comment / favorite |
| Schedule | Unified scheduled-trigger entry for all three app types · 6 schedule modes + state machine + run history |
| Workspace | Jump entry · Switch to Layer 2 to enter the workspace the user manages |
Layer 2 · Workspace (build & govern side)
Who it's for: workspace administrators + app builders — product managers, developers, business operators.
Core action: create + orchestrate + govern. Assemble AI capabilities inside the workspace they own.
| Resource category | Capabilities |
|---|---|
| Apps | Create prompt / orchestrated Agents · Configure tools · Set permissions · Publish to Workbench |
| Apps | Drag-and-drop Workflow orchestration · Triggers · Version management |
| Data | Knowledge base (RAG) · Data source (SaaS / database connector) |
| Capabilities | Tools (MCP / HTTP plugin / EvoTool marketplace) · Skills |
| Workspace management | Members & permissions · Workspace-level observability · Settings |
Layer 3 · Organization Management (system administrators)
Who it's for: organization administrators, IT leads, security leads.
Core action: org-level governance. Manage the global resources, people, models, and security that sit above all workspaces.
8 Module Groups
| Group | Modules | Purpose |
|---|---|---|
| Org structure | Members & departments · Roles & permissions · Organization info | People management + RBAC role definitions + organization metadata |
| Model services | Model definition · Model interface platform · Model deployment · Default model configuration | Global model governance: custom models / interface routing / deployment management / global defaults |
| Tools & skills | System tools · Organization tools · Organization skills · Web search | Org-level Tool / Skill registration + web search provider configuration |
| API management | API Key | API keys exposed by the organization (for third parties calling the Evose platform API) |
| Invitations | Invite members · Invite to workspace | Invite organization members + invite to workspaces |
| Analytics | Observability | Three pillars (Logs / Metrics / Traces) + 4 dimensions (Organization / Workspace / Resource / User) |
| Security | Keys & credentials · Resource policy | Third-party API credentials (used by Evose to call external services) + resource access policies |
| Desktop client | Security policy · Security audit | Security configuration for the desktop client + audit logs (supported in both Private and SaaS) |
→ Enter Organization Management
Relationships Between the Three Layers
Users may have cross-layer identities: a product manager is a builder in Workspace A (Layer 2) and a Workbench user calling other people's Agents (Layer 1). Each layer's permissions are controlled by RBAC + ACL.
SaaS vs Private — Differences at Each Layer
| Layer | SaaS | Private |
|---|---|---|
| Layer 1 Workbench | Identical | Identical |
| Layer 2 Workspace | 1 organization holds N workspaces | 1 organization holds N workspaces |
| Layer 3 Org management | Multi-tenant, multi-organization · Platform billing (Credits) | 1 deployment = 1 organization · No platform billing; customer-configured models |
Next Steps
- Want each layer's specific capabilities → Use / Build / Govern
- Want a technical layering view → Core capabilities / Technical architecture