Agentic AI concepts
Practical guides to the ideas, practices, protocols and standards behind AI agents in production.
- Harness engineering: the system around the modelA model is not an agent. The harness around it — loop, tools, permissions, guardrails and evaluation — decides whether it works in production.
- Context engineering: what the model sees, and whenMost agent failures are context failures. Context engineering is the discipline of deciding what the model sees at every step.
- Agentic RAG: retrieval that the agent controlsClassic RAG retrieves once and hopes for the best. Agentic RAG turns retrieval into a tool the agent plans, repeats and checks.
- Evaluating AI agents and RAG: evals, metrics and RAGASIf you can't measure an agent, you can't improve it, release it safely or defend it in an audit. Evals are how you measure.
- AgentOps: running AI agents in productionShipping an agent is the start, not the end. AgentOps is the operational discipline that keeps agents reliable, affordable and accountable after launch.
- AI agent protocols: MCP, A2A, AG-UI, A2UI, AP2 and UCPAgents talk to people, to other agents, to tools and to markets. Each of those layers now has its own open protocols. Here is the map.
- The OWASP Top 10 for Agentic Applications, explainedAgents don't just say things — they do things. The OWASP Agentic Top 10 lists the ten ways that goes wrong. Here they are in plain language, with controls.
- Governing AI agents with ISO/IEC 42001 and the NIST AI RMFAgents act with delegated authority, so governance has to reach past documents into running systems. Two established frameworks show how.