What is an AI agent?
An AI agent is an LLM-based system that decides its own steps and uses tools to complete a task, such as investigating a security alert end to end.
Definition
AI agent
Also called Agentic AI, LLM agent, Autonomous agent
An AI agent is a system in which a large language model dynamically directs its own process and tool use to complete a task, deciding which steps to take rather than following a fixed, predefined script.
How it works
Anthropic groups LLM systems that use tools under the umbrella of agentic systems, and draws a line between workflows, where LLMs and tools are orchestrated through predefined code paths, and agents, where the LLM dynamically directs its own process and tool use. Its basic building block is an LLM augmented with retrieval, tools, and memory, which can generate its own search queries, choose tools, and decide what to retain.
An agent runs in a loop: read the task, decide on a next step, call a tool (run a query, fetch a file, call an API), read the result, and repeat until it can answer or needs a human. Tools are often connected through standards like the Model Context Protocol. Anthropic recommends starting with the simplest solution that works, noting that agentic systems often trade latency and cost for better task performance.
OWASP's Agentic Security Initiative points out that integrating LLMs has expanded agents' scale, capabilities, and risks. An agent with tools can be manipulated through prompt injection, can misuse permissions it was given, and can act on a hallucinated conclusion.
Example
A triage agent receives an alert for a suspicious OAuth app consent. It looks up the app's publisher and permissions, queries sign-in logs for the user, checks whether other users granted the same app, and finds that a single user consented from their usual device to a well-known productivity app. It closes the alert with that evidence attached. Had the app requested mailbox access and come from an unverified publisher, it would have escalated with the same summary.
Why it matters for a SOC
Agents can do the repetitive context-gathering behind most alerts, which is why they underpin the AI SOC. The questions that matter are control and evidence: which model, which tools, which instructions, and which actions need approval. Cotool's AI SOC lets teams control the prompt, model choice, tools, and output format for each agent, stores agents and skills in Git as YAML with review and rollback, and measures them with evals.
Related terms
- Model Context Protocol (MCP)The Model Context Protocol (MCP) is an open standard that defines how AI applications connect to external tools, data sources, and workflows, so a model can read context and take actions through a common interface.
- Prompt injectionPrompt injection is an attack in which text supplied to a large language model, directly by a user or hidden in content the model reads, alters its behavior in ways its developers didn't intend, such as ignoring instructions, leaking data, or misusing tools.
- Alert triageAlert triage is the first review of a security alert, in which an analyst or automated system confirms what happened, judges whether it is malicious and how severe it is, and decides to close it, escalate it, or start a response.
- AI hallucinationAn AI hallucination is output from a generative AI model that is false, fabricated, or inconsistent with its inputs, but presented confidently as if it were accurate.
Go deeper
- What is an AI SOC?An AI SOC uses AI agents to triage, investigate, and close or escalate security alerts with evidence. How it works, how it differs from SOAR, and the risks.
- How to evaluate an AI SOC platformTest AI SOC vendors on your own closed alerts, check the evidence behind each verdict, and review data access, models, permissions, and pricing.
FAQ
- What is the difference between an AI agent and a workflow?
- Anthropic distinguishes the two: workflows orchestrate LLMs and tools through predefined code paths, while agents let the LLM decide its own steps and tool use. Workflows are more predictable for well-defined tasks, and agents suit tasks that need flexibility.
Sources
- Building effective AI agents | Anthropic · accessed November 16, 2026
- Agentic AI - Threats and Mitigations | OWASP Gen AI Security Project · accessed November 16, 2026
- AI SOC Without the Black Box | Cotool · accessed November 16, 2026
