# CURA Capability Catalog

## Identity and platform structure
- Full-stack web application used to create, manage, orchestrate, and monitor AI agents
- Organizational hierarchy for deployment: Organization / Department or Collective / Team / Agent
- Multi-tenant platform model documented in internal system materials
- Microsoft Entra ID based authentication and identity model documented in platform materials

## Governance and oversight
- Human-in-the-Loop Review (HILR) for sensitive operations
- Audit trails and operational logging
- Role-based and approval-led governance model
- Human reviewer gates embedded into high-trust workflows

## Memory and continuity
- Institutional Brain with six scopes: Organization, Collective/Department, Team, Agent, User, Session
- Agent Notebook for structured entries separate from semantic memory
- Reusable Skills library for scoped instruction sets

## Agent execution and coordination
- Individual agents for specific roles
- Multi-agent team orchestration
- Scheduling for one-time, interval, and recurring tasks
- Agent-to-agent communication documented in platform materials

## Tools and integrations
- Microsoft Graph tools for Outlook email, calendar, Teams, and SharePoint
- Built-in web search
- Built-in file search
- Built-in code interpreter
- MCP server support
- Custom API tools

## Workflow examples grounded in current materials
- Private-Sector Contribution Verification workflow:
  - Lead Scout
  - Verification Officer
  - Record Builder
  - Three human gates
  - No agent performs the final submission to the external reporting system
- Mission-driven operating use cases described publicly:
  - Grant operations
  - Donor stewardship
  - Intake and safeguarding
  - MEL reporting

## Scope and roadmap boundaries
- Documented live platform architecture includes governance, memory, scheduling, Microsoft-native integrations, and extensibility
- Internal product direction also describes future connector ecosystem expansion, OpenAPI import, OAuth connection management, and partner/private connector models
- Public site does not yet fully expose all of these details in structured machine-readable form