AI Agent
An AI agent is a software system built on a language model that perceives context, chooses actions from a toolset, and executes them in a loop to reach a goal.
In plain English
An AI agent is a loop wrapped around a language model. On each turn it looks at the conversation so far — the goal the user set, past tool results, its own plan — and decides what to do next. The options are usually "call a tool," "answer the user," or "stop." Tools are typed functions the agent can invoke: a web search, a file write, an email send, an API call. The model picks the tool, fills in the arguments, and waits for the result before its next turn.
That loop is what separates an agent from a chatbot. A chatbot produces text. An agent produces actions: files created, rows updated, deploys triggered, messages sent. Good agents also keep short-term memory of what they tried, what worked, and what failed, so they don't repeat themselves across turns. The frontier agent frameworks — Claude Agent SDK, OpenAI's tool use, LangGraph — all center on this same read-decide-act-observe pattern.
Why it matters for Black Box
Black Box is built from agents. A CEO agent coordinates work, and 18 specialist agents — covering coding, content, research, browser, business ops, design, product, marketing, sales, social, SEO, ads, PR, customer success, support, data, finance, and legal — execute it. Every action the product takes — shipping a landing page, scoring a lead, drafting a newsletter — is an agent running this loop on your behalf.
Examples
- A research agent that reads three sources, synthesizes a briefing, and writes it to a file.
- A coding agent that reads a repo, writes a patch, runs the tests, and opens a pull request.
- An inbox triage agent that labels mail, drafts replies, and escalates anything it is not confident about.