AxioRankThe Security Gateway for AI Agents
Open the console

Glossary

The language of AI agent security.

Clear, self-contained definitions of the terms that matter once your AI can take actions. From prompt injection and MCP to guardrails and zero trust for agents.

Start freeSee the platform

Plain language · accurate · no signup to read

Agent concepts

The building blocks: what an AI agent is, and how it reaches the outside world.

AI agent

An AI agent is a system built on a large language model that can take actions in the world by calling tools, not just generate text.

Read the definition

Agentic AI

Agentic AI describes systems that plan, decide, and act over multiple steps toward a goal, rather than responding to a single prompt.

Read the definition

Model Context Protocol (MCP)

The Model Context Protocol is an open standard that lets AI agents connect to external tools and data sources through a common interface.

Read the definition

MCP server

An MCP server is a program that exposes tools, resources, and prompts to AI agents over the Model Context Protocol.

Read the definition

Tool call

A tool call is the moment an AI agent invokes an external function, with concrete arguments, to take an action or fetch data.

Read the definition

Attacks and threats

How AI agents get compromised, from prompt injection to poisoned tools.

Prompt injection

Prompt injection is an attack that embeds malicious instructions in text an AI reads, causing it to ignore its original instructions and follow the attacker's.

Read the definition

Indirect prompt injection

Indirect prompt injection hides malicious instructions in third-party content an agent reads, such as a web page, email, or tool result, rather than in the user's own message.

Read the definition

Jailbreak (LLM)

A jailbreak is a prompt crafted to bypass a language model's safety training so it produces output it was designed to refuse.

Read the definition

Data exfiltration (AI agents)

Data exfiltration is when an AI agent is manipulated into sending sensitive data, such as secrets or personal information, to a destination the attacker controls.

Read the definition

Tool poisoning

Tool poisoning hides malicious instructions inside a tool's name or description so the AI model is manipulated the moment it reads the tool, before any call is made.

Read the definition

Tool shadowing

Tool shadowing is when a malicious tool impersonates or overrides a trusted one, so the agent calls the attacker's version instead.

Read the definition

MCP rug pull

An MCP rug pull is when an MCP server behaves safely during review, then silently changes its tool definitions to malicious ones after it has been approved.

Read the definition

Agent kill chain

An agent kill chain is a sequence of individually plausible tool calls that together carry out an attack, such as reading a secret and then sending it out.

Read the definition

Defense and governance

The controls that keep an autonomous agent inside its lane.

AI agent security gateway

An AI agent security gateway sits on the boundary between an agent and its tools, scoring, enforcing policy on, and recording every tool call.

Read the definition

AI guardrails

Guardrails are the runtime checks that constrain what an AI agent can say or do, blocking or reshaping unsafe inputs, outputs, and actions.

Read the definition

Zero trust for AI agents

Zero trust for AI agents applies the never trust, always verify principle to agents: every agent, tool, and call is authenticated and authorized, nothing is trusted by default.

Read the definition

Human in the loop

Human in the loop is a control where a person must approve a high-risk agent action before it runs, keeping a human accountable for consequential decisions.

Read the definition

AI red teaming

AI red teaming is the practice of deliberately attacking an AI system, with prompt injections, jailbreaks, and adversarial inputs, to find weaknesses before real attackers do.

Read the definition

Excessive agency

Excessive agency is the risk that an AI agent is granted more capability, permission, or autonomy than its task requires, widening the damage it can do if compromised.

Read the definition

Blast radius (AI agents)

Blast radius is the extent of harm an AI agent could cause if it were compromised, determined by the data it can reach and the actions it can take.

Read the definition

Definitions are the start. Enforcement is the point.

AxioRank scores every tool call your agents make for leaked secrets, PII, destructive operations, and prompt injection, checks it against your policy, and proves it in a tamper-evident audit log. Start free, no card.

Start freeTalk to us