AI agent
An AI agent is software that uses an AI model to work toward a goal by taking actions: it plans steps, calls tools such as search, code, email or a CRM, looks at what happened and decides what to do next. A chatbot answers a question. An agent goes and does the thing.
Why it matters
Agents move AI from advice to execution. A support agent can look up the order, issue the refund and reply to the customer. A coding agent can read the codebase, make a change and run the tests. That saves far more time than a good answer does, and it also means the agent’s mistakes now have real effects.
What an agent needs
The model supplies the reasoning, but most of what makes an agent useful sits around it, in the agent harness:
- Tools it can call, often connected over MCP.
- Context about the task, the company and what has already been done.
- Permissions that limit what it can read and change.
- A stopping rule, including when to ask a person instead of guessing.
When an agent fails, the cause is usually one of these, not the model. It didn’t have the information, it had access it shouldn’t have, or nobody told it when to stop.
Agents and other agents
Once everyone on a team runs their own agent, the agents start to overlap. Two of them edit the same file, or one repeats work another already did. Agents that can see each other’s progress, hand off tasks and flag conflicts avoid most of this. Agents that work in isolation leave the coordinating to the people.
How elvex relates
In elvex, each person’s agent joins a shared thread acting for that person, sees only what they can see, and works alongside teammates’ agents: claiming and handing off to-dos, flagging conflicts and asking the right person when a decision is needed.