Implementation Guide¶

This guide provides step-by-step instructions for implementing the RMACD Framework in your organization.

Prerequisites¶

Before implementing RMACD, ensure you have:

1. Inventory of AI agents — Document all deployed or planned autonomous agents
2. Identified stakeholders — Change management, security, compliance teams
3. Approval workflow infrastructure — Ticketing system, notification capabilities

Step 1: Choose Your Implementation Model¶

Two-Dimensional Model (RMACD + HITL)¶

Choose this if: - Your organization does not have formal data classification - You need rapid adoption - You're piloting with a limited scope

Three-Dimensional Model (RMACD + HITL + Data Classification)¶

Choose this if: - You have an established data classification scheme - You operate in a regulated industry - Your agents access data across multiple sensitivity levels

Step 2: Define Permission Profiles¶

Start with the provided templates:

• Observer — Read-only agents
• Logistics — Read + Move agents
• Provisioning — Read + Move + Add agents
• Operations — Read + Move + Add + Change agents
• Administrator — Full RMACD (with restrictions on Restricted data)

Customize as needed for your agent types.

Step 3: Configure Policy Enforcement¶

Integrate RMACD with your agent runtime:

1. Load permission profiles from your policy store
2. Intercept agent operation requests
3. Evaluate against the governance matrix
4. Enforce autonomy requirements (allow, notify, queue for approval, deny)
5. Log all decisions for audit

Step 4: Integrate Approval Workflows¶

Map autonomy levels to your existing systems:

Step 5: Test and Validate¶

Before production enforcement:

1. Run in audit-only mode (log decisions, don't enforce)
2. Review logs for unexpected denials or approvals
3. Adjust profiles and matrix as needed
4. Gradually enable enforcement

Step 6: Monitor and Iterate¶

Ongoing operations:

1. Review agent behavior patterns
2. Adjust permissions based on demonstrated trustworthiness
3. Respond to incidents and update profiles
4. Conduct periodic compliance audits

Python Tools Registry¶

A reference Python implementation is available in the tools-registry/ directory:

tools-registry/
├── rmacd_tools_registry.py       # Core implementation (750+ lines)
├── example_usage.py              # Usage examples with 27 pre-configured tools
├── test_registry.py              # Test suite (43 tests)
├── mcp_integration.py            # MCP auto-classification bridge
├── rmacd_tools_catalog.json      # Pre-configured tool catalog
├── rmacd_permission_profiles.json # Standard permission profiles
└── README.md                     # Detailed documentation

Key features: - Tool registration with RMACD classification - Permission validation against agent profiles - Risk scoring for tools and workflows - Audit logging for compliance - MCP tool auto-classification

See tools-registry/README.md for detailed usage instructions.

Platform-Specific Guides¶

Coming soon:

• Kubernetes / Container orchestration
• AWS / Azure / GCP agent integration
• LangChain / AutoGPT integration
• ServiceNow integration

Need Help?¶

• Email: contact@rmacd-framework.org
• Web: rmacd-framework.org
• GitHub Discussions: rmacdframework/spec/discussions