Agents and protocols
How agents connect to tools, applications, and other agents—and where protocols end and application responsibilities begin.
6 published articles
Articles
September 14, 2026
The Agentic AI Foundation: What Its Projects Do
A practical tour of the Agentic AI Foundation's projects and their roles in tools, repository instructions, agent collaboration, and infrastructure, with a dated look at governance and maturity.
August 18, 2026
AGENTS.md: Writing Useful Instructions for Coding Agents
Build an AGENTS.md around a real repository task. Write concrete instructions, choose their scope, check how Codex or Claude Code loads them, and evaluate the result before adding more guidance.
August 16, 2026
The Agent Works Like a New Hire
A capable assistant can investigate broadly while its tools limit what it may commit. An invoice and a coding workflow show where reusable tools, automatic approvals and human judgment fit, and why approval fatigue deserves more than another button.
August 11, 2026
AI Agent Architecture: A Replenishment System Example
Follow a nightly replenishment agent from inventory analysis to buyer approval. Work through its permissions, interface, MCP and A2A integrations, data destinations, and a printable architecture worksheet.
August 4, 2026
AG-UI and A2UI: Connecting an Agent to Your Application
Follow a support-ticket request through AG-UI, then render an A2UI interface using an existing component library. See how context, state, controls and user decisions fit together.
July 28, 2026
WebMCP: When a Page Declares Its Actions
How does an AI assistant carry out a request on a website? WebMCP lets the page describe actions it can perform. Follow an inventory change from the user's request through the browser to the page's code and back to the assistant.