Agents

AI agents are starting to act on our behalf: answering questions, preparing meetings, and soon taking actions. Each one is only as useful as what it knows about the person it works for, and today that knowledge mostly lives inside whichever company runs the agent.

Writing on this topic looks at what agent memory actually is, who holds it, and what happens when someone wants to use more than one agent, or to stop using one altogether. It covers the difference between memory assembled from what someone typed and a record of what actually happened, and why that difference matters more as agents take on real work.

It also covers how agents reach the information they need. Protocols such as MCP let an agent connect to outside tools and data, which makes it possible for a person to keep their own record and let any agent draw on it, only as far as they allow.

The question running through all of it is simple: when an agent knows you well, whose knowledge is that?