Agent Memory
Short-term context, long-term memory, retrieval, state stores, summarization, compaction, and forgetting policies.
Core area
AI Agents
Agent memory, long-running workflows, durable harnesses, tool boundaries, evals, replay, and governance.
Thesis
Useful agents are not prompts in a loop. They need memory policy, bounded execution, tool mediation, durable state, approval points, replay, evals, and clear failure handling.
Core areas
Long-Running Agents
Durable execution, checkpoints, resume, async work, approval pauses, and failure recovery.
Agent Harnesses
Planning loops, tool mediation, skills, hooks, MCP configs, cross-harness support, and security boundaries.
Tool Use and MCP
Tool schemas, auth boundaries, MCP servers, approval gates, sandboxing, and audit logs.
Evals and Replay
Run traces, eval workflows, replay, policy checks, cost/latency monitoring, and PII/safety controls.
Enterprise Deployment
Azure AI Foundry, Agent Framework, backend runtimes, SSE, ops consoles, observability, and role-based workflows.
Projects and artifacts
af-agent-harness
Agent harness work around planning loops, tool mediation, and deployable agent architecture.
PythonAzure Agent FrameworkMCP
What it demonstrates: A practical harness layer for agent workflows beyond single prompt chains.
af-durable-harness-examples
Examples for durable agent execution, checkpoints, resume behavior, and long-running workflows.
PythonDurable workflowsAgents
What it demonstrates: How agent runs can be structured as recoverable workflows.
af-pii-multi-agent
Multi-agent prototype focused on PII-aware processing and governance boundaries.
PIIGovernanceAgents
What it demonstrates: How sensitive-data controls can be modeled before agents call tools or external systems.
agent-framework-ozg
Workshop-ready Agent Framework fork with runnable samples for agents, workflows, memory, reasoning, and Azure AI Foundry.
PythonAzure AI FoundryAgent Framework
What it demonstrates: How agent framework examples can move from learning material into prototype and workshop assets.
aviation-demos-01
Aviation demo workspace for domain-specific AI patterns.
AviationDomain AIPrototype
What it demonstrates: How domain constraints shape AI product architecture.
emergency-payment
Emergency payment processing demo with sanctions screening, liquidity assessment, and operational procedure agents.
FastAPINext.jsAzure AI Foundry
What it demonstrates: How regulated workflows need tool boundaries, streaming status, evidence trails, and explicit handoff points.
everything-claude-code
AI coding workflow system with skills, agents, MCP configuration, and guardrail conventions.
Agent harnessMCPDeveloper tooling
What it demonstrates: How agent harnesses can use skills, memory, review loops, and security workflows as an operating surface.
finance-ontology-demo
Finance ontology demo for structuring domain knowledge and AI workflows.
Knowledge graphFinanceAI workflow
What it demonstrates: Ontology-backed domain AI system design.
foundry-demo
Hands-on Azure AI Foundry workshop covering hosted agents, tracing, governance, grounding, MCP tooling, and A2A flows.
Azure AI FoundryOpenTelemetryMCP
What it demonstrates: How enterprise agent deployments need identity, observability, grounding, tool governance, and deployment repeatability.
graph-ai-tutor
Local tutoring app prototype using graph-backed learning state, API services, and web UI.
GraphTutoringNode
What it demonstrates: How domain AI products can combine graph structure, local state, and guided learning loops.
tradingagents-agent-harness-demo
Domain agent harness demo using trading-style research and workflow patterns.
AgentsFinance demoWorkflow
What it demonstrates: Domain-specific agent orchestration and tool-mediated analysis.
treasury-shock-day-demo
Banking and treasury operations demo for multi-agent liquidity, FX stress, approval, audit, and compliance workflows.
Domain AITreasuryMulti-agent
What it demonstrates: How domain workflows require traceability, policy context, approval points, and role-aware agent coordination.