A Practical Introduction to Agentic Coding

Why It Matters

Agentic coding transforms developers from code writers to orchestrators, unlocking faster delivery and higher code quality while reshaping the software development workflow.

Key Takeaways

• Agentic coding gives AI full workspace access for autonomy
• GitHub Copilot’s agent mode can execute terminal commands directly
• Providing context files improves code quality and consistency significantly
• Custom agents can be built using SDKs like LangChain framework
• Agentic loops consist of context gathering, tool action, verification, and feedback

Summary

The ACM Tech Talk introduced “agentic coding,” a new paradigm where AI agents act as autonomous programmers within a developer’s environment. Hosted by Abigail Misy Dobe and presented by Microsoft senior developer advocate Marlene, the session highlighted how open‑source communities and professional bodies are promoting these tools.

Marlene traced the evolution from simple code‑completion to chat‑based LLMs and finally to true agentic coding. In VS Code’s GitHub Copilot “agent mode,” the model receives full workspace context, can run terminal commands, and even debug its own output, turning a passive suggestion engine into an active coding partner.

Live demos showed practical techniques: attaching GitHub issues, pull‑requests or example files as context, using standardized instruction files (agents.md) to steer behavior, and building a custom agent with the Copilot SDK and a lightweight GPT‑4.1 model. The audience saw the agent greet them and respond to prompts, illustrating the end‑to‑end loop of gathering context, taking action, and verifying results.

The emergence of multiple coding agents—Copilot, Claude Code, CodeX, and DIY frameworks like LangChain—creates a competitive ecosystem that promises higher productivity but also demands disciplined context management. Developers who master the agentic loop can accelerate feature delivery, reduce manual debugging, and shape the next generation of AI‑augmented software development.

Original Description

Title: A Practical Introduction to Agentic Coding

Date: March 11, 2026

Duration: 1HR

ABSTRACT

In this session, Marlene Mhangami shares how she gets the most out of agents in her development workflow using GitHub Copilot SDK and GitHub CLI. She walks through how she uses MCP (Model Context Protocol), Agent Skills and Instructions to create semi-autonomous agents that can complete multi-step tasks end-to-end.

SPEAKER

Marlene Mhangami

Microsoft Python & AI Expert

MODERATOR

Abigail Mesrenyame Dogbe

Open Source Researcher and Programs Manager