> ## Documentation Index > Fetch the complete documentation index at: https://docs.myrmex.ai/llms.txt > Use this file to discover all available pages before exploring further. # The AI Agents > Meet the team of specialized AI agents you direct from the Workspace. Centurion coordinates the specialists; you steer them with @-mentions and approve what they do. import { CardGroup, Card, Note, Tip } from '@mintlify/components'; Myrmex is run by a team of specialized **AI agents** you talk to from the [AI Workspace](/documentation/ai-workspace/overview). Each agent is an expert in one area — endpoints, integrations, threat intelligence, posture, reporting — and they work together on your behalf. You describe what you need in plain language, and the right specialists take it from there. **"Agent" means two different things in Myrmex.** On this page, an *agent* is an **AI specialist** you chat with. That is separate from the **software agent** — the program installed on a host (see [Agent capabilities](/documentation/agent-endpoint-security/features)). The AI agents *operate* the software agent; they are not the same thing. ## How You Direct the Team You steer the team with `@`-mentions. There is **no model or agent selector** — you simply mention the specialist you want: * **Mention an agent** — type `@` and pick a specialist (for example, `@Perseus`) to send your request straight to them. * **Mention no one** — [Centurion](/documentation/multi-agent-system/agent-orchestrator), the orchestrator, reads your request and routes it to the right specialist (or several) automatically. You can also `@`-mention **context entities** — a device, integration, alert, report, or finding — to scope the work to exactly what you mean. ## AI Proposes, You Approve The agents investigate and analyze freely, but anything that **changes your environment** — isolating a host, applying a fix, running a command, scheduling an update — is proposed for your review first. You stay in control: the AI does the heavy lifting, and you approve what actually runs. ## The Roster The orchestrator and your single point of contact. Interprets your request and routes it to the right specialists — including its SOC mode for alerts and investigations. The endpoint manager. Tracks the health, configuration, and installed software of the assets you've registered in Myrmex. The endpoint specialist. Investigates hosts and runs response actions — isolate, quarantine, kill a process, remediate, harden. The integration specialist. Operates your connected tools across network devices, security software, databases, and cloud platforms. The threat intelligence analyst. Investigates indicators of compromise, domains, IP reputation, and adversary infrastructure. The security posture strategist. Scores your environment against frameworks like CIS, NIST, and ISO, and prioritizes what to fix first. The researcher. Gathers open-source intelligence and external context from across the internet. The author. Turns analysis into professional reports, policies, and audit documents ready to share. ## Working Together The real power shows when a request needs more than one specialist. Say you ask Myrmex to investigate a suspicious alert on a server: 1. **Centurion** interprets the request and builds a plan. 2. **Orion** enriches the indicators — checking domains, IPs, and reputation. 3. **Perseus** inspects the host itself, looking at the processes and connections behind the alert. 4. **Athena** checks whether the host's posture left an opening worth fixing. 5. **Scribe** writes up the findings as a shareable report. You watch it happen in the Workspace, and you approve any action that changes a system before it runs. New to the platform? Start with [Core concepts](/documentation/get-started/concepts) to see how the agents fit into contexts and the Workspace, then try the [Quickstart](/documentation/get-started/quickstart).