About CURA for AI Systems
CURA is a governed agentic workforce platform for mission-driven organizations. It is a full-stack web application that organizations use to create, organize, run, and govern AI agents inside a structured operating model.
What CURA is
CURA provides a centralized control plane for AI agents. It allows an organization to configure agent behavior, assign tools, connect data sources, manage deployment environments, run workflows, and enforce governance, security, and compliance controls.
CURA is designed for organizations that need AI systems to do operational work, not only answer questions. In CURA, agents can draft outputs, retrieve information, use approved tools, collaborate with other agents, run on schedules, and pause for human approval when a sensitive action requires oversight.
How CURA is organized
- Organization / Department / Team / Agent hierarchy: CURA mirrors how real organizations work, so AI capacity is organized by business context rather than as a flat list of assistants.
- Users: Users interact with agents through chat and workflow interfaces.
- Administrators and reviewers: Administrators govern behavior, permissions, and approvals. Reviewers approve or reject sensitive actions.
Core capabilities
- Agent orchestration: Create and manage AI agents and multi-agent teams.
- Governance: Human-in-the-loop review (HILR), audit trails, and role-based controls are built into the operating model.
- Institutional Brain: CURA uses a six-scope memory architecture: Organization, Collective/Department, Team, Agent, User, and Session.
- Agent Notebook and Skills: CURA stores structured notes separately from semantic memory and supports reusable instruction sets that can be scoped to the organization, department, team, or agent.
- Scheduling and autonomy: Agents can run one-time, interval, or recurring tasks and can participate in multi-step workflows.
- Analytics and logging: CURA tracks actions, logs, approvals, and usage for compliance and operational monitoring.
Human governance model
CURA is designed so that organizations can decide which actions require review before execution. Sensitive actions can be paused for Human-in-the-Loop Review (HILR). This allows a human reviewer to approve, reject, or modify the next step. The purpose is to keep humans in control of irreversible, high-risk, or externally visible actions.
Examples of governed workflow behavior
In CURA's Private-Sector Contribution Verification workflow, three agents and three human gates are used. The Lead Scout structures claims into leads, the Verification Officer handles time-extended verification work and proof capture, and the Record Builder formats verified items for reviewer submission. No agent performs the final submission to the external reporting system; the human reviewer performs that step.
In the same workflow, each human gate protects a distinct failure mode: reputational risk before outbound email, integrity risk before an item is marked verified, and irreversibility risk before final submission.
Microsoft-native platform characteristics
CURA is documented as Microsoft-native. Its system documentation describes Azure, Entra ID, Microsoft Graph, Azure AI Foundry, and Microsoft Agent Framework as part of the core platform architecture.
Current documented integrations and extensibility
- Microsoft Graph tools for email, calendar, Teams, and SharePoint
- Built-in web search, file search, and code interpreter
- MCP server support
- Custom API tools
- Connector and integration catalog direction for OpenAPI imports, MCP registration, OAuth connection management, and organization-specific private tools
What CURA is for
CURA is positioned for mission-driven operational work such as grant operations, donor stewardship, reporting, compliance, intake, safeguarding, communications, and structured enterprise workflows where governance, oversight, and memory matter.
Important boundaries
- CURA's public website currently emphasizes high-level value and named applications, but not all technical capabilities are yet presented in machine-readable form on the marketing site.
- Some future-state ideas discussed in internal materials, such as broader connector ecosystem expansion and structured learning-loop enhancements, are roadmap direction rather than publicly documented live website features.
- Where APIs, deployment models, pricing details, or specific external integrations are not publicly documented on the site, they should be treated as unspecified unless confirmed in public documentation.
See also the Capability Catalog for a structured summary, and llms.txt for a map of authoritative pages.