> ## Documentation Index
> Fetch the complete documentation index at: https://docs.amnify.ai/llms.txt
> Use this file to discover all available pages before exploring further.
# Set up a penetration test
> Authorize an application, verify the domain, and choose how deep to go.
Amnify runs real penetration tests against applications you own and explicitly authorize. A team of agents probes the live application the way an attacker would, and nothing is reported unless it was actually exploited.
## Applications and targets
A **pentest application** is your declaration that you may authorize active testing. Under it you define **targets**: the host and endpoints to test. Optionally attach API documentation, credentials, and source repositories.
## Domain verification is mandatory
A target cannot run until you prove you control the domain. Amnify gives your organization a unique value. Publish it as a DNS TXT record on the name it shows, or as plain text at the well known URL it shows, then check.
The value is unique to your organization, so proof of control cannot be transferred. A run against an unverified target is refused.
## Choosing the run
Two independent choices.
| Engagement | What Amnify has |
| ------------- | -------------------------------------------------------------------------------- |
| **Black box** | The authorized host and endpoints only |
| **Gray box** | The above, plus a credential, so testing is authenticated |
| **White box** | The above, plus your source, so findings trace to code and fixes can be proposed |
The highest impact, most commonly exploited ways in: broken authentication and access control, injection, remote code execution, server side request forgery, exposed secrets. Light traffic, low cost. It will not chain small issues into one attack path.
Audit ready. Everything in Quick plus the full exposed surface, per role access control, business logic and race conditions, file upload abuse, prompt injection, and chained attack paths. Every severity, at higher traffic and cost.
## How a run works
A run moves through four phases: surface mapping, discovery, validation, and, in a white box run, a fix that is re-tested.
Only validated results become findings: a working, minimal, non destructive proof of concept, or for a known vulnerable dependency, its version independently confirmed as present. In a white box run, Amnify also patches the cause and re-runs the proof to show it is gone.
## Scheduling
Daily, weekly or monthly, per combination of engagement and package. A target can hold up to six schedules, one per combination, each pausable.
Penetration tests and scans never overlap on the same scope. Different scopes can run at the same time.
Each finding carries the vulnerability class, affected endpoint or package, CVSS vector and score, CWE, the proof of concept, and the engagement and package that produced it. Full method: [How a pentest runs](/pentest/how-it-works).