AI Agent

An AI agent is a language model wrapped in a control loop that lets it use tools (search, code execution, APIs), observe their output, and decide what to do next. The model is the brain; the tools are the hands; the loop is what makes it 'agentic'.

Unlike a single-shot chatbot reply, an agent can take dozens or hundreds of steps to complete a task. Modern coding agents like Claude Code and Cursor's Composer can plan a refactor, execute shell commands, run tests, and iterate until the job is done. Customer-support agents like Intercom Fin can read a knowledge base, ask clarifying questions, and resolve tickets without a human in the loop.

The quality of an agent depends on three things: model reasoning, tool design, and loop reliability. The frontier in 2026 is making agents trustworthy enough to run unattended for long horizons.