Claude for Wireless Core Operations and 5G Core NOC Automation

1. Claude Basics and the Wireless Core NOC Mission

• Claude orientation and Claude Chat fundamentals.
• Using Claude to support technical analysis and operations teams.
• Using Claude to explain telecom concepts and summarize operational problems.
• Initial AI safety boundaries and read-only limits in production networks.
• The role of a NOC in a 24x7 firefighting environment.
• Why 5G Core operations are complex.
• Overview of AMF, SMF, UPF, PCF, UDM, AUSF, NRF, gNB, UE, PDU sessions, QoS, and MTTR.
• Introduction to the five-layer agentic AI architecture.
• Introduction to the local Open5GS or free5GC architecture.
• Hands-on exercise: use Claude to create a 5G Core glossary.
• Hands-on exercise: verify the local open-source 5G Core sandbox.
• Hands-on exercise: summarize the Wireless Core firefighting problem and draft the five-layer architecture map.

2. Prompt Engineering and Early Governance for Fault Triage

• Role prompting and context setting.
• Structured JSON and Markdown output.
• Diagnostic prompting and escalation logic.
• Adding LOW_RISK and HIGH_RISK classifications to technical recommendations.
• Alarm interpretation and symptom-versus-root-cause analysis.
• Severity classification and evidence gathering.
• Integrating standard operating procedure guardrails into diagnostic workflows.
• Hands-on exercise: create prompts for KPI review, fault classification, and runbook lookup.
• Hands-on exercise: produce standardized risk level, human approval, and rollback metadata.
• Hands-on exercise: test prompts against high latency, data access failure, and call-drop incidents.

3. Wireless Core Data Sources and Local Telemetry Extraction

• Technical document and operational log analysis.
• Requirements extraction and runbook drafting.
• Creating service-isolation tables.
• Preparing raw files for ingestion by local tools.
• Identifying active AMF, SMF, and UPF logs, subscriber database status, and network traces.
• Identifying critical indicators for Data, Voice, 911, and Messaging services.
• Analyzing delay, low speed, latency, packet loss, and chronic degradation indicators.
• Hands-on exercise: locate and extract diagnostic logs and trace files from the local 5G Core.
• Hands-on exercise: use Claude to draft operational runbooks from local evidence.
• Hands-on exercise: build a machine-readable data-source inventory and service-isolation matrix.

4. Claude Code, Local MCP Connectors, and Shared Workspace

• Claude Code installation and terminal configuration.
• Designing custom local MCP servers with Node.js or Python.
• Exposing local files, log directories, status scripts, and database state through MCP tools.
• Separating read-only tools from write-action scripts.
• Telemetry ingestion and event normalization.
• Protecting high-risk operational scripts with local approval gates.
• Hands-on exercise: install and configure Claude Code in the local sandbox.
• Hands-on exercise: build a local MCP server that retrieves 5GS logs and database state.
• Hands-on exercise: configure MCP safety policies that disable destructive tools.

5. Claude Code Foundations and 5GS Wrapper Walkthrough

• Reading and tracing unfamiliar codebases with Claude Code.
• Using Claude to explain system APIs.
• Safe local code modification and small refactoring.
• Verifying changes with local unit tests.
• Open-source 5GS script-wrapper topology.
• Programmatic network-function states: Active, Offline, and Out of Rotation.
• Telemetry exporters, severity metrics, and event loggers.
• Hands-on exercise: inspect the 5GS control scripts from the command line.
• Hands-on exercise: generate a visual architecture map of the codebase.
• Hands-on exercise: create a script wrapper that queries subscriber session states.

6. KPI Monitoring and Real-Time Fault Detection

• Writing lightweight dashboard scripts or UI components with Claude Code.
• Converting raw telemetry streams into real-time metrics.
• Debugging state transitions and event logs.
• Monitoring active UEs, throughput, latency, packet loss, and MTTR.
• Simulating high latency, low throughput, packet loss, and chronic degradation.
• Designing the KPI Monitoring Agent behavior model.
• Hands-on exercise: build a Python or Node.js script that polls local 5GS metrics.
• Hands-on exercise: feed metrics to Claude Code through an MCP tool to trigger detection alerts.
• Hands-on exercise: simulate a chronic UPF degradation loop and verify automatic detection.

7. Use-Case Agents and Wireless Core Specialist Agents

• Multi-agent role design and responsibility scoping.
• Implementing specialist-agent prompts with controlled tool access.
• Managing safe handoffs between specialist scripts.
• Defining escalation boundaries when a specialized tool fails.
• Designing KPI Monitoring, Fault Detection and Isolation, RCA, Remediation, and Guardrail and Risk agents.
• Designing AMF, SMF, UPF, PCF, AUSF and UDM, IMS and Voice, Messaging, and gNB domain specialists.
• Introducing the NEXUS Supervisor routing model.
• Hands-on exercise: write the underlying prompts for each specialist agent.
• Hands-on exercise: build a file-based message queue for diagnostic handoffs.
• Hands-on exercise: create routines that retrieve telemetry for Data, Voice, 911, and Messaging faults.

8. NEXUS Supervisor, Critical Service Isolation, and RCA Flow

• Writing supervisory orchestration logic in local scripts.
• Creating dynamic routing tables from incoming JSON telemetry.
• Implementing fallback strategies when script tasks fail.
• Generating chronological Markdown audit logs of agent decisions.
• Supervisory routing, severity classification, and priority queues.
• Critical service isolation for Data, Voice, 911, and Messaging.
• Root-cause analysis correlation loops.
• Hands-on exercise: build the NEXUS Supervisor event-processing loop.
• Hands-on exercise: implement immediate priority escalation for 911 and IMS events.
• Hands-on exercise: generate automated RCA reports in a local output folder.

9. Guardrails, Security, and Human-in-the-Loop Node Control

• Reviewing and auditing previously implemented prompt guardrails.
• Implementing interactive CLI approval prompts for high-risk tools.
• Testing edge cases to verify that unauthorized scripts are blocked.
• Designing state-rollback workflows.
• Enforcing Layer 5 guardrails and execution limits.
• Node Out-of-Rotation decision paths.
• Comprehensive audit logging and security boundaries.
• Hands-on exercise: require human confirmation before restarts or Out-of-Rotation actions.
• Hands-on exercise: build an automatic rollback mechanism for failed recovery actions.
• Hands-on exercise: conduct red-team tests using unauthorized destructive-command requests.
• Hands-on exercise: verify that local security profiles block prohibited actions.

10. Capstone Simulator Demo and Copilot Enterprise Discussion

• Orchestrating the complete local 5G Core NOC simulation.
• Validating the Detection → Triage → RCA → Remediation → Guardrail Check → Approval → Action → Verification workflow.
• Simulating real-time critical faults and KPI drops.
• Simulating high latency, packet loss, and random call drops.
• Simulating UE registration and PDU session failures.
• Simulating IMS call setup failures and call drops.
• Simulating mobile-originating and mobile-terminating messaging failures.
• Simulating 911 service-routing and availability concerns.
• Simulating node Out-of-Rotation recommendations and human approval.
• Generating incident summaries, post-mortems, and shift handoff reports.
• Hands-on exercise: run the complete simulator against the local open-source 5G Core.
• Hands-on exercise: demonstrate human-in-the-loop validation for node Out-of-Rotation decisions.
• Hands-on exercise: generate a polished shift handoff report and executive incident summary.
• Strategic discussion of Claude Code and Microsoft 365 Copilot enterprise collaboration.