> ## 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.

# Integrations Overview

> Connect your security stack to Myrmex — how integrations work, how credentials stay safe, and how to add one.

import { CardGroup, Card, Note, Steps, Step, Tip } from '@mintlify/components';

Integrations connect Myrmex to the tools you already run — cloud providers, firewalls, network gear, EDR/XDR, SIEMs, databases, ticketing, and more. Once connected, the AI agents (led by [Hydra](/documentation/multi-agent-system/integration-specialist-agent)) can read data from and take action on those systems on your behalf, right from the [Workspace](/documentation/ai-workspace/overview).

## How Integrations Work

Every integration is an **instance** you create from a connector, scoped to the [context](/documentation/get-started/concepts#how-your-account-is-organized) you're working in. Depending on the system, Myrmex connects in one of a few ways:

<CardGroup cols={2}>
  <Card title="Through a Collector" icon="server">
    For on-premises and network devices, a **Myrmex Collector** (a device running
    the agent in Collector mode) reaches the system over SSH or its API from
    inside your network.
  </Card>

  <Card title="Directly to the cloud" icon="cloud">
    For cloud and SaaS platforms, Myrmex connects over the vendor's API — via
    OAuth consent or a least-privilege service credential.
  </Card>
</CardGroup>

<Note>
  **Your credentials never pass through the AI.** The agents only ever reference
  an integration by its ID and ask it to run a specific action; the actual
  secrets are stored securely and resolved server-side. See
  [Data Protection](/documentation/management/data-protection).
</Note>

## Shared vs. Private Credentials

By default an integration uses **one shared credential** for the whole organization — everyone works through the same service account. For integrations where each person should act as themselves, you can switch the integration to **private**: every user links their **own** credential, and any request runs under the credential of the user performing it.

<CardGroup cols={2}>
  <Card title="Shared (default)" icon="users">
    One organization-wide credential. Anyone with access uses the same service
    account — simplest for centrally-managed systems.
  </Card>

  <Card title="Private (per user)" icon="user-lock">
    Each user links their own login. Actions are attributed to — and use the
    permissions of — the individual running them. *"Requires each user to use their
    own credentials."*
  </Card>
</CardGroup>

**Turning on private mode.** On the integration's detail view, an administrator (with the `integrations.update` permission) flips **Private integration**. It's only available for connector types that support private credentials; a shared integration keeps working exactly as before.

**Linking your account.** When an integration is private, each user links a personal credential once:

<Steps>
  <Step title="Create a personal credential">
    Under **Settings → My Accounts**, add an account — your personal login for a
    private integration (API key, basic auth, OAuth, SSH, or database credential).
  </Step>

  <Step title="Link it to the integration">
    Open the integration and go to **Integration Details → Private credential →
    My credential**, choose your account, and **Link** it. You can create a new
    account inline here too.
  </Step>

  <Step title="Check the status">
    Your linked credential shows as **Valid**, **Invalid**, or **Not tested** — a
    per-user status, independent of the integration's overall health.
  </Step>
</Steps>

<Note>
  Until you link a valid credential, a private integration appears **unavailable to
  you** and the AI can't act on it on your behalf — you'll be prompted to *"link
  your account in Integration Details → Private credential."* Other users are
  unaffected; each person's access is their own.
</Note>

## What You Can Do with an Integration

* **Read and analyze** — pull alerts, inventory, configuration, and metrics into your investigations, in real time and on demand.
* **Take action** — apply firewall rules, isolate endpoints, run queries, open tickets, and more, always with the [AI-proposes-you-approve](/documentation/get-started/concepts#ai-proposes-you-approve) safeguard for changes.
* **Measure posture** — run continuous and on-demand [Security Posture](/documentation/security-posture/overview) audits against the connected system.

## Adding an Integration

<Steps>
  <Step title="Open Add Integration">
    From the **Directory**, choose the **Add Integration** quick action.
  </Step>

  <Step title="Pick the connector">
    Choose the system you want to connect from the catalog.
  </Step>

  <Step title="Provide the credential">
    Follow that connector's page in this section to create the required credential
    (API key, service account, SSH user, or OAuth consent) and fill in the fields.
  </Step>

  <Step title="Connect">
    Save to validate the connection. The integration then appears under
    **Environment → Integrations**.
  </Step>
</Steps>

## The Connector Catalog

Browse connectors by category in the sidebar. Each connector has its own page describing exactly which credential to create and how.

* **Cloud & CSPM** — AWS, Google Cloud, Oracle Cloud, Prowler
* **Microsoft & Identity** — Entra ID, Intune, Microsoft 365, Sentinel, Google Workspace
* **Firewalls & WAF** — FortiGate, Palo Alto, Sophos, pfSense, WatchGuard, Cisco, Juniper, Barracuda, Cloudflare, Azion
* **Network Devices** — Cisco, Huawei, Juniper, MikroTik, Dell, Aruba, D-Link, 3Com, Intelbras, Meraki, UniFi, Umbrella, NetBox
* **EDR & XDR** — CrowdStrike, SentinelOne, Trend Micro Vision One, Bitdefender
* **SIEM & Monitoring** — Wazuh, Elastic Security, IBM QRadar, Rapid7 InsightIDR, Zabbix, FortiAnalyzer
* **Vulnerability & Threat Intelligence** — Tenable, Google Threat Intelligence, Shodan, Apura, Axur
* **ITSM & Ticketing** — GLPI, Jira
* **DevOps, Automation & Notifications** — GitHub, Kubernetes, Grafana, n8n, SendGrid
* **Databases** — PostgreSQL, MySQL/MariaDB, SQL Server, Oracle, Firebird, MongoDB, Redis, Cassandra, Elasticsearch

<Tip>
  Need something not listed? Many devices can still be reached over **generic SSH**, and the connector catalog is continually expanding. Ask in the Workspace what's available for your environment.
</Tip>

## Connecting Your Own Accounts

Separately from these organization-wide integrations, you can link **personal connectors** — Gmail, Google Drive, Outlook, and SharePoint — so the AI can work with documents from your own accounts. These act as you, using per-user consent.