MG/AI Hub
Agentic engineering, in plain language.
Agent orchestration
Agent orchestration is coordinating several AI agents, and the people overseeing them, so that a multi-step piece of work is planned, divided, executed and verified reliably.
Agent skills
Agent skills are reusable, written instructions that teach an AI agent how a team performs a specific task, such as shipping a feature, reviewing security or writing release notes.
Agentic engineering
Agentic engineering is building software with AI agents that plan and carry out whole tasks, such as changing code, running tests and opening pull requests, under human oversight.
Agentic SDLC
The agentic SDLC is a software development life cycle in which AI agents carry out stages such as analysis, design, implementation, testing, review and release, with humans approving key decisions.
AI agent
An AI agent is a software system, usually built on a large language model, that can pursue a goal by deciding on steps and using tools such as code editors, terminals or APIs.
AI code review
AI code review uses an AI model or agent to examine code changes, usually in a pull request, and flag bugs, security issues and style problems before or alongside human reviewers.
AI guardrails
AI guardrails are the technical and process controls, such as permissions, approvals, tests and monitoring, that keep AI agents operating safely within agreed limits.
Human-in-the-loop
Human-in-the-loop means designing AI workflows so that people review and approve important decisions, such as plans, code merges and releases, before they take effect.
Model Context Protocol (MCP)
The Model Context Protocol (MCP) is an open standard that lets AI applications connect to external tools and data sources, such as ticketing, design or browser tools, through a common interface.
Sub-agent
A sub-agent is a specialised AI agent started by another agent to handle one part of a larger task, with its own instructions, tools and permissions.