Agentic Coding for C# Developers using Codex

Install the codec cli globally with npm, log in using a chatgpt account, and trust the directory to start codecs, choosing a subscription for payment.

Install the Codex extension in Visual Studio Code, set the chat sidebar beside Co-Pilot, sign in to chat-tbt (or use an API key), and prepare to start your first prompt.

Install the Codex app from OpenAI developers, via Windows or Mac, preferably from the Microsoft Store; then log in, choose a ChatGPT subscription, and add a Codex project.

Explore codecs configurations across the CLI, IDE, and app, including model and mode, plus full access permissions. Learn to manage MCPs, plugins, and controls like startup and steer versus queue.

Explore a Blazor server to-do app built with ModBlazor and a shared API. See SQLite with EF Core powering repository patterns, task CRUD, and an MCP server.

Explore basic prompting for an existing to-do app, adding on-page stats and a home.json target, with three sections for this month, last month, and total, in agentic coding.

Learn about the evolving agents.md standard, how rules guide coding agents across Codex, Cursor, GitHub Copilot, and Claude Code, and manage root versus multiple agent files with the .agents file.

Translate codex-guided agents.md rules into an editor config file and enforce errors, not warnings, so AI and humans follow coding standards in C# projects.

Examine rules beyond the agents.md file, covering architecture, api endpoints, and Blazor page creation, and explain how agent skills handle conditional rules separate from the rules you always should follow.

Discover AgentSkills, the open standard adopted by the agentic community, and learn how skills stored as folders in .agents use skills.md with YAML front matter to load on demand.

Learn how prompts leverage skills by default, how an agent reads relevant skill guides, and use commands like add Blazor page or add stats page.

Demonstrate building a Microsoft To Do integration using the Graph API, covering Azure app registration and task.read/task.write permissions, and a JavaScript-based CRUD skill with environment variables.

Assess whether to reference skills in AgentsMD and MCP tools; avoid duplicating skill definitions, since an AI agent will read AgentsMD and all skills natively.

Automate checks to keep agents.md and skills up to date without rewriting, fix spelling and version errors, and reflect changes in the readme and generated code for consistency.

Learn how MCP servers standardize tools and data access for ai agents. Explore information-based and action-based servers, plus local and remote deployments, with Codex as a client.

Set up MCP servers in Codex using GUI, CLI, or config toml, choosing local stdio or streamable remote servers with URLs, bearer tokens, headers, and environment variables.

Discover recommended MCP servers for C# development, including Microsoft Learn, Context 7, GitHub, and Azure DevOps, and learn to build your own lightweight MCP solutions using Codecs and internal tools.

This lecture reviews the April 17, 2026 Codex update, covering bug fixes, a non-project chat, in-app image generation, and new project tabs for summary, review, and change files.

Examine how to set reasoning effort levels when coding with Codex, comparing medium and high to extra high, and when to use high for coding vs non-coding tasks.

Learn how to use requestUserInput to present options in a GUI from prompts, enabling normal mode prompts with a GUI picker by enabling the defaultModeRequestUserInput feature flag in config.toml.

Start a cloud task in Codecs, apply its change to a red button, and see it locally in Visual Studio, with an option to revert.