AI agents, explained
AI agents, explained.
An AI agent is a model that keeps going after one answer: it remembers, follows a written method, uses your tools and asks before it does something it cannot undo. A chatbot answers once. An agent finishes the job, because the parts around the model do the work.
- 01
A model that keeps going until the job is done.
What is an AI agent?
- 02
The model is the engine. The harness is the car.
What is an agent harness?
- 03
It remembers Friday on Monday.
AI agent memory
- 04
Write the method once. Anyone's agent can run it.
AI agent skills
- 05
The plug, not the appliance.
What is MCP?
- 06
It can draft anything. It sends only when you say so.
AI agent permissions
- 07
Work that starts without anyone asking.
Scheduled AI agents
- 08
Everyone's agent, in one room.
Multiplayer AI for teams
- 09Coming soon
A safe room for the code it writes.
Agent sandboxes
- 10Coming soon
It reads and clicks websites for you.
Browser use
- 11Coming soon
It uses a whole desktop, the way you do.
Computer use
- 12Coming soon
It works where your team already talks.
Channels
- 13Coming soon
More agents isn't always better.
Agents working together
- 14Coming soon
How you know it works.
Evals
Card 02 is open today, and it is the picture every other card hangs on. The rest land in order, each one written around the same small case: a ten-person accounting firm and the one report it sends every Friday.
Related: what multiplayer AI means for a team