From Prompts to AI Agents | Foundry Agent Service Explained
summary
Explains how Foundry Agent Service runs multiagent systems with managed compute, state, security and a unified tool catalog, reducing build time from months to days; demonstrates C# and Python SDKs, native connectors for SAP and Salesforce, and memory tiers (short, long, entity, quality state) that resume crashed workflows; warns about production risks like infinite-loop runaway, latency from per-step reasoning and flaky tool APIs, and shows orchestration patterns (manager-worker, peer‑to‑peer) and built‑in guardrails. Foundry Agent Service SDK native persistence and observability unified tool catalog
AI-written summary · about
chapters (22)
- 0:00Introduction
- 1:45Shift to autonomous agents
- 4:12What is an AI agent
- 5:57Four stages of autonomy
- 6:38AI agents in supply chain
- 8:02What makes an agent smart
- 9:26Four pillars of architecture
- 10:50Agent reasoning loop
- 11:51Foundry Agent Service engine
- 12:53Foundry vs DIY
- 13:56Platform orchestration tools
- 14:59Blueprint for agents
- 16:01Agent memory deep dive
- 17:23Multi-agent patterns
- 18:04Task orchestration logic
- 18:47End-to-end architecture
- 19:27Production risks
- 20:49Monitoring and debugging
- 21:51Guardrails and security
- 22:33Future of agentic AI
- 23:18Foundry SDK overview
- 24:21Q&A
related talks
22mBuild and Manage AI Apps w/ Your Agent Factory
05h 17mBuilding AI Applications
58mAdvanced Autonomous AI Agents: Memory, Collaboration and Evolution in Action
56mBuilding AI Agents and Multi-Agent Systems with Azure AI
20mGenAI agents - Evolution of automation
26mAI powered automation & multi-agent orchestration
24mAI Agents in Production: Why the Platform Matters
39mEvolving Intelligence: Designing Autonomous AI Agents