A six-part Python+Agents series teaches building agents and AI-driven workflows with the Microsoft Agent Framework: tool integrations, RAG and Redis/Mem0 memory, OpenTelemetry observability, Azure AI Evaluation SDK, multi-agent orchestration, and human-in-the-loop checkpoints.
We just concluded a six-part livestream series that maps building AI agents and workflows in Python. The series covers agents, memory, observability, workflows, orchestration, and human-in-the-loop controls.
Main feature/change and impact
The Microsoft Agent Framework provides a unified Python toolkit for building agentic systems. It standardizes agent anatomy, tool calling, and supervisor patterns for multi-agent coordination. The framework integrates context via RAG with databases and long-term memory stores like Redis or Mem0. OpenTelemetry tracing and the Azure AI Evaluation SDK enable production-grade observability and measurable quality. This reduces integration effort and increases reliability for complex agent workflows.Practical implications
Developers can prototype end-to-end agent workflows using provided demos and slides. Code samples support frontier LLMs and Foundry models for realistic testing. Workflows support executors, edges, events, conditional branching, and structured outputs to avoid brittle string checks. Multi-agent patterns include handoff and planning supervisors, enabling parallel execution and result aggregation. Human-in-the-loop features allow request-and-response checkpoints and tool approval for safer operation.“We just concluded Python + Agents, a six-part livestream series where we explored the foundational concepts behind building AI agents in Python using the Microsoft Agent Framework.”The series delivers recordings, slide decks, and open-source code repositories for immediate use. Join weekly office hours on Foundry Discord to ask technical questions and iterate on agent designs. Prioritize instrumenting agents with OpenTelemetry and running automated evaluation with Azure AI Evaluation SDK. Next steps: clone the python-agentframework-demos repo, run examples against your preferred model endpoints, and add RAG and Redis memory to validate behavior across sessions.
Key points from the article:
Related Coverage:
- From apps to agents: Rearchitecting enterprise work around intent
- Agents League: Meet the Winners
- Copilot in the C-Suite: Infobip’s Veselin Vuković on using Copilot to nurture partnerships
From the Microsoft Developer Community Blog articles
