November 18, 2026
Pentesting in 3 Easy Steps: How VeritySec Puts AI Agents to Work on Your Security
VeritySec
By Neville
9 min read
Penetration testing is valuable, but traditional approaches can be expensive, time-consuming, and difficult to repeat consistently.
A typical engagement involves coordinating security professionals, defining the scope, running reconnaissance, analyzing results, manually investigating potential vulnerabilities, and eventually producing a report. Depending on the complexity of the environment, this process can take weeks.
Automated vulnerability scanners solve part of the problem. They can examine large attack surfaces quickly and identify known vulnerabilities, but speed doesn't necessarily translate into useful security intelligence. A scanner can produce hundreds of findings without establishing which ones are exploitable, which matter in the context of the application, or what an organization should prioritize.
The result is familiar: lengthy reports, false positives, disconnected evidence, and findings that developers struggle to act on.
AI introduces another possibility.
What if security professionals could work alongside AI agents capable of planning assessments, operating real security tools, investigating results, and maintaining a record of their activities?
Not another chatbot that explains cybersecurity concepts. Not another scanner that dumps CVEs into a report.
An AI-driven platform that can actually participate in the security testing process.
That's the idea behind VeritySec.
What Is VeritySec?
VeritySec is an open-source, AI-driven security testing platform that brings AI agents and real security tools into a managed environment.
The central idea is straightforward: instead of manually coordinating every step of an assessment, you interact with specialized AI agents through a conversational interface. These agents can plan activities, select appropriate tools, execute tasks, analyze results, and maintain context throughout an engagement.
Think of it as having an AI-assisted security team working under your supervision.
You begin by creating a project and registering the assets you want to assess. These might include IP addresses, domains, or web applications.
From there, you can start a conversation with a specialized security agent. The agent can use available tools to investigate the target, explore its attack surface, and identify potential weaknesses.
The platform is designed to connect the entire process:
- AI-assisted reconnaissance and security testing.
- Real security tools rather than simulated command execution.
- Specialized agents for different security tasks.
- Human approval before potentially dangerous operations.
- Evidence collection and vulnerability tracking.
- Project organization, access control, and engagement monitoring.
And getting started takes three basic steps.
Step 1: Bring Your Own API Key
VeritySec follows a Bring Your Own Key (BYOK) model.
Rather than forcing users into a proprietary AI subscription or restricting them to one model provider, the platform allows them to configure their own API credentials.
Supported providers include OpenAI, Anthropic, OpenRouter, and compatible APIs.
API credentials are encrypted at rest using AES-256 and resolved by the backend when requests are made.
Why does this matter?
1. Control over costs
You pay your chosen AI provider directly. You aren't tied to an additional platform-level AI markup or an opaque token pricing structure.
Your usage and costs remain associated with your provider account.
2. Freedom of model choice
Different models have different strengths.
You might prefer one model for planning and reasoning, another for analyzing technical output, and a third for reviewing findings.
With BYOK, you're not locked into a single provider's ecosystem. You can choose the models that best suit your workflow and budget.
This flexibility is particularly important in AI-driven security testing, where model capabilities, reasoning quality, context limits, and costs can significantly affect the results.
VeritySec provides the orchestration layer. You retain control over the AI providers you use.
Step 2: Open a Conversation With a Specialized Agent
VeritySec isn't designed around a single general-purpose AI assistant.
It provides specialized security agents, each configured for a particular type of work.
These include:
- Pentester: General penetration testing and security assessment.
- Forensics Analyst: Investigation and analysis of digital evidence.
- Cloud Auditor: Assessment of cloud environments, including AWS.
- CTF Solver: Assistance with capture-the-flag challenges and security exercises.
Instead of starting from scratch every time, you can select an agent suited to the task and begin a conversation.
The conversational interface is more than a convenient way to submit commands. It provides continuity.
You can return to an engagement and continue working with its existing context, including previous activities, findings, and investigation progress.
This changes the interaction model.
Traditionally, a security professional might move between terminals, scripts, scanners, spreadsheets, and reporting tools to coordinate an assessment.
With VeritySec, the goal is to bring much of that workflow into one place.
More Than One Agent
Some security assessments involve several different types of work.
Reconnaissance, vulnerability investigation, exploitation validation, and reporting require different approaches.
VeritySec also supports multi-agent workflows, allowing specialized agents to participate in more complex tasks.
The broader idea is to move beyond treating AI as one assistant that must do everything.
Instead, different agents can contribute their particular capabilities to a coordinated assessment.
This is an important direction for AI-driven security tools: specialization and coordination rather than simply making one chatbot larger.
Step 3: Provide Your Target
This is where the actual security testing begins.
Register your authorized target, such as a domain, IP address, or web application, and give the agent its instructions.
Depending on the engagement and the tools available, the agent can work through several stages.
1. Reconnaissance
The agent can investigate the target's exposed attack surface using tools for activities such as:
- Port scanning.
- Subdomain enumeration.
- Service identification.
- Technology fingerprinting.
- Web crawling and endpoint discovery.
The objective is to understand what is exposed before deciding where to investigate further.
2. Vulnerability Discovery
Once the initial attack surface has been mapped, the agent can investigate potential weaknesses.
This may involve checking for known vulnerabilities, examining application behavior, and using tools such as Nuclei to identify potential security issues.
It can also use specialized tools for particular classes of vulnerabilities, depending on the target and assessment requirements.
3. Exploitation and Validation
Finding a potential vulnerability is not the same as proving that it exists.
A scanner might identify a potentially vulnerable service, but further investigation may be necessary to establish whether the issue is actually exploitable.
Within an authorized testing environment, VeritySec can use tools such as SQLMap and other security utilities to investigate suspected vulnerabilities.
The goal is to move from an initial observation toward a finding supported by reproducible evidence.
4. Evidence Collection
Security findings are only useful when they can be investigated and acted upon.
VeritySec is designed to associate findings with supporting evidence, including raw tool output, screenshots, and other relevant observations.
Findings can also be organized using established security classifications, including OWASP categories and MITRE ATT&CK mappings.
This makes it easier to understand not just what the agent reported, but why the finding matters and what evidence supports it.
The Human Still Has the Final Say
There is an important distinction between an AI agent that can execute commands and one that can be trusted to execute them without supervision.
Penetration testing involves operations that can disrupt services, modify data, trigger security alerts, or affect systems outside the intended scope.
Giving an AI agent unrestricted access to security tools is therefore not a sensible default.
VeritySec incorporates a human-in-the-loop approval mechanism.
When an agent attempts an operation that requires approval, execution pauses. The user can review the proposed action, its parameters, and the intended target before deciding whether to approve or deny it.
This is particularly important for potentially destructive, intrusive, or noisy operations.
For example, an agent might determine that it needs to perform an active exploitation test. Instead of automatically proceeding, it can request permission.
The user remains responsible for the decision.
This serves two purposes.
First, it introduces an additional layer of control over automated security operations.
Second, it makes the process observable. Users can examine what the agent intends to do and review the tools it selects throughout the engagement.
For someone learning penetration testing, this can also be valuable. Rather than simply reading a tutorial explaining a particular methodology, they can observe an AI agent attempting to apply it to a real, controlled environment.
The objective isn't to remove human involvement. It's to make human involvement more informed and effective.
What Makes VeritySec Different?
There are already plenty of vulnerability scanners, security automation platforms, and AI-powered chat interfaces.
So where does VeritySec fit?
1. It's Not Just Another Vulnerability Scanner
Traditional scanners are useful for identifying potential security issues, but their output often requires substantial investigation.
VeritySec introduces an AI reasoning and orchestration layer on top of real security tools.
An agent can interpret results, decide what to investigate next, and use additional tools based on what it has discovered.
An open port, for example, might lead to service identification, which could lead to further investigation of the running service.
The important difference is the attempt to connect individual tool executions into a broader assessment workflow.
2. It's Not Just a Chatbot
A conversational interface alone doesn't make a platform a security testing tool.
VeritySec agents can orchestrate actual utilities such as Nmap, Nuclei, and SQLMap.
When an agent invokes one of these tools, the platform executes the corresponding utility within its managed execution environment and processes its output.
The intention is to make AI an active participant in security testing rather than simply a source of recommendations.
3. It Emphasizes Traceability
Security testing needs an audit trail.
Organizations need to understand what was tested, which operations were performed, what was discovered, and who authorized particular actions.
VeritySec includes project organization, vulnerability tracking, engagement reporting, and role-based access control.
Its RBAC system supports four roles:
- Admin: Platform and administrative access.
- Operator: Performs security testing activities.
- Auditor: Reviews activities and findings.
- Viewer: Has access to permitted information without operational privileges.
Resource-scoped permissions provide additional control over access to projects and their associated resources.
This is particularly relevant when security testing involves multiple people or teams.
4. You Retain Control of Your AI Providers
The BYOK approach gives users flexibility over their model providers and associated costs.
Rather than depending on a single proprietary AI service, users can configure compatible providers and choose the models they want to work with.
Combined with its open-source approach, this gives users more control over how they deploy and operate the platform.
Under the Hood: The Architecture
For those interested in the technical implementation, VeritySec is built around a Go backend and a lightweight frontend.
The backend handles application logic, AI orchestration, tool execution, project management, and other platform services.
Docker provides an isolated environment for executing security tools.
The platform integrates more than 60 security tools, covering different areas of reconnaissance, vulnerability discovery, and security testing.
These are actual tools, not AI-generated simulations.
When the agent runs Nmap, it invokes Nmap. When it uses Nuclei, it executes Nuclei templates against the authorized target.
The platform manages tool availability, execution, and output handling so that agents can interact with the tools through a consistent interface.
VeritySec also integrates with Burp Suite, allowing security professionals to incorporate the platform into an existing web application testing workflow.
The architecture is intended to bring AI orchestration, security tooling, and engagement management together without requiring users to coordinate every component independently.
Who Is VeritySec For?
VeritySec is being developed for people and teams who want to explore a more integrated approach to security testing.
Some potential use cases include:
Independent penetration testers
Solo security professionals can use AI agents to assist with reconnaissance, investigation, and repetitive assessment activities.
Security teams
Teams can organize projects, manage assets, track vulnerabilities, and maintain records of security testing activities.
Developers
Developers can use the platform to investigate the security of applications they own or are authorized to test.
CTF participants
Security learners and competition participants can experiment with specialized AI assistance while solving challenges.
Security trainers
Trainers can demonstrate how security testing workflows operate and use AI agents to make the process more interactive.
The platform isn't intended to replace experienced security professionals. Its purpose is to help people organize and automate parts of the work while retaining control over consequential decisions.
The Bigger Idea: Where Does AI Fit Into Penetration Testing?
The interesting question isn't simply whether AI can run Nmap or execute a vulnerability scanner.
Those capabilities already exist.
The more important question is whether AI agents can coordinate the different stages of a security assessment, understand the results they receive, make sensible decisions about what to investigate next, and produce findings that can withstand independent verification.
That is a considerably harder problem.
An AI agent can misunderstand a tool's output. It can pursue an irrelevant lead, overlook an important vulnerability, or report an issue that cannot be reproduced.
Running a tool successfully doesn't automatically mean the assessment was successful.
This is why evidence, scope enforcement, human approval, and independent verification matter.
We believe the next generation of security testing platforms will be shaped by how well they combine these elements.
The objective isn't to give an AI unrestricted access to every security tool available.
It's to build systems in which AI can perform useful work, explain its actions, preserve evidence, and remain accountable to the people supervising it.
That's the direction we're exploring with VeritySec.
Getting Started
VeritySec is an open-source project, and the basic workflow is designed to be straightforward:
- Create an account and access the platform.
- Add your API key in the settings.
- Create a project and register your authorized target.
- Start a conversation with a specialized security agent.
- Review its proposed activities and approve operations when required.
- Follow the assessment and investigate the resulting findings.
Always test systems you own or have explicit authorization to assess. Define the engagement scope before running security tools.
Final Thoughts
AI is changing how software is built, how infrastructure is managed, and how technical work is performed.
Cybersecurity won't be an exception.
We are moving toward a world where AI agents can do more than explain commands or generate scripts. They can interact with tools, investigate systems, coordinate workflows, and assist with increasingly complex technical tasks.
But capability alone isn't enough.
The real challenge is building systems that can do useful work without sacrificing control, traceability, and trust.
VeritySec is our attempt to explore that possibility: an open-source platform where AI agents work with real security tools, under human supervision, with an emphasis on evidence and accountability.
We're building toward a future where security professionals spend less time coordinating tools and more time making informed security decisions.
VeritySec is an open-source project. Its security testing capabilities should only be used against systems you own or have explicit authorization to assess.