This guide takes you from an empty profile to your first AI Network Pentesting assessment. By the end you'll have a scope in place, a validated configuration, and an assessment running against your network.

## Prerequisites

**Your firewall rules.** If a firewall or IPS sits in front of your targets, decide whether to allowlist Escape's [public IP addresses](<https://docs.escape.tech/documentation/platform/firewall/#public-locations>). Allowlisted traffic shows what an attacker gets past your filters; blocked traffic shows what your filters stop.

## Create a Profile

Go to **AI Network Pentesting \> New Profile**. The form has four steps.

### 1\. Scope

Enter one target per line: an IPv4 address, an IPv4 CIDR range, an IPv6 address, or a domain. To import a list, drop a CSV file on the field or click **Import CSV**.

```text
edge.example.com
192.0.2.10
192.0.2.0/24
2001:db8::1
```

The form shows what it parsed: the **IP ranges**, the **Domains and addresses**, the total number of IPv4 addresses, and any invalid lines. The first domain or single address becomes the profile's main host, and the profile name is generated from it.

A scope needs at least one domain or single address. See [Scope](<https://docs.escape.tech/documentation/network-pentesting/scope/index.md>) for every rule and limit.

*The **Scope** step parses each line into IP ranges and domains or addresses, then the **Attack vectors** step lists what agents may attempt.*

### 2\. Attack Vectors

Select what the agents are allowed to attempt. You need at least one vector.

- **Safe** vectors only observe: they list services, read configuration, and check TLS, without signing in.
- **Intrusive** vectors try credentials, send injection payloads, exploit known CVEs, or send repeated requests. They can lock accounts or disturb services, so select them only with written authorization.

### 3\. Fine-Tune (Optional)

- **Context**: describe the scope in **What is this scope?** (for example, "Corporate edge and VPN"), and list what matters most. Agents use this to prioritize. It narrows their focus, it never widens the scope.
- **Location**: where traffic comes from. Pick the Europe, United States, or Canada region, or one of your [private locations](<https://docs.escape.tech/documentation/platform/private-location/index.md>).
- **Duration**: the rate limit and the maximum duration. See [Guardrails](<https://docs.escape.tech/documentation/network-pentesting/guardrails/index.md>).
- **Schedule**: start now, or pick a later start time.

### 4\. Review &amp; Launch

1. Click **Validate Configuration**. The check runs from the selected location and doesn't start the pentest.
2. Fix any target the check flags, then validate again.
3. Click **Create &amp; run assessment**, review the summary in the confirmation dialog, and confirm.

**Create &amp; run assessment** stays disabled until validation passes. To keep the profile without running it, click **Save profile**.

## Follow the Assessment

Open **AI Network Pentesting \> Assessments** and select your run.

- **Reasoning** shows the agents live: what they're testing, the tools they call, and what they find.
- **Issues** fills up as findings are reproduced.
- **Assets** lists the hosts and services the assessment saw.

To stop the run, cancel it from the assessment page. See [Results](<https://docs.escape.tech/documentation/network-pentesting/results/index.md>) for how to read each tab.

## Next Steps

- Tune the [Guardrails](<https://docs.escape.tech/documentation/network-pentesting/guardrails/index.md>) before you test fragile or production systems.
- Run an [AI Pentesting](<https://docs.escape.tech/documentation/ai-pentesting/quickstart/index.md>) assessment on the web apps and APIs your network assessment exposed.