Editorial pick · we run itDecima Systems, our editorial pick for AI-driven application testing, is run by the publisher of this site. Everything else is compared on what it publishes. DisclosureRev 2026-11-08
Pentesting ToolsCatalogue of AI security testing software

Method · rev 2026-11-08

How the catalogue is built

In brief

Every entry is built from the maker’s own site, repository or pricing page, linked and dated. Open-source tools are ordered by GitHub stars; commercial tools by how much they publish. There are no scores or star ratings, and the editorial pick, which we run, is a judgment kept outside every list.

01

What gets in

A tool qualifies if it is software that does security testing with an AI model, falls in one of three lanes, and has public pages we can cite.

  1. Autonomous and AI-driven platforms

    45 entries. Commercial software that runs penetration tests itself: against web applications and APIs, or against networks, Active Directory and cloud accounts. Sold as SaaS, an on-premises appliance or an agent you deploy.

  2. AI red-teaming tools

    35 entries. Software that attacks LLM applications, models and AI agents: prompt injection, jailbreaks, data leakage, unsafe tool use. Open-source kits and commercial platforms.

  3. Open-source AI pentest agents

    20 entries. Code you run yourself, with your own model keys, that uses an LLM to plan and carry out offensive testing: of web apps, networks or capture-the-flag targets.

Left out: consulting firms whose AI-assisted offer is an engagement delivered by their staff; plain vulnerability scanners and DAST tools with no AI-driven testing product; code-review tools; and open-source projects with under about 1,000 GitHub stars, unless a major security publication has covered them (the entry says so). Abandoned projects, with no commits for 18 months, are dropped.

02

What we record

  1. What it tests

    Web apps, APIs, networks, Active Directory, cloud, mobile, LLM apps, AI agents, code: only what the maker’s own documentation says.

  2. How it runs

    SaaS, on-premises, a CLI or Docker container, a Python library, an MCP server.

  3. How findings are checked

    Whether the maker says a person reviews findings, a second agent reproduces them, or nothing is said. Quoted short, never upgraded.

  4. Price and free use

    The published price exactly as written, the unit it is sold in, and any free tier or trial. Nothing estimated: an unpublished price reads “Not published”.

  5. Open-source facts

    GitHub stars, licence, latest release and last commit, read from the GitHub API on the check date. Stars change daily; the date is on every entry.

  6. Ownership and cautions

    Acquisitions and renames from the companies’ own announcements; safety or trust concerns only from named, reputable sources, linked.

This revision: 100 tools, 634 distinct source URLs, and published prices for 39 vendors. Euro prices are converted at 1.1165 USD per euro (ECB reference rate 2026-11-06).

03

Order, and why there are no scores

Open-source tools are listed by GitHub stars on the check date, highest first. Stars measure attention, not quality, but they are public, dated and the same for everyone.

Commercial tools are listed by how many of four things their own pages publish: a price, a free tier or trial, what the product tests, and how findings are checked. Ties go alphabetically. That rewards transparency, which a buyer can use, and says nothing about testing quality, which we cannot see.

We don’t publish ratings or scores. A score from us would be an opinion presented as a measurement. Judge a tool on your own target, with written authorisation, or ask the vendor for a sample report.

Entries carry a lane code such as P·04. It is the entry’s position in this revision’s order, used for reference. It is not a rank.

04

The editorial pick

Decima Systems is labelled Editorial pick · we run it wherever it appears. The pick is a judgment, not a measurement, and it is made by the owner of the firm being picked: the publisher of this site runs Decima. That is why it is shown in its own box and kept out of every ordered list and comparison table.

Its entry follows the same evidence rule as everyone else’s: facts from its public site, dated, with its pilot, planned and in-development labels kept. Its prices are re-read from decima.systems before each update. More on theabout page.

05

Upkeep and corrections

This revision was checked on November 8, 2026. Prices, stars and products change often (one vendor in this catalogue repriced in July 2026, another runs a promotion that ends in December), so confirm with the maker before you buy. The pricing research behind /pricing/ was first compiled for Decima’s own pricing work; only the vendors’ published prices are used here.

If an entry about your product is wrong or out of date, emailengage@decima.systems with a link to the public page that shows the right fact. We correct entries from public sources only, and no one can pay to change one.

06 · sources

Sources for this page

  1. GitHub REST API: repositories — Where stars, licences, release and commit dates are read from.
  2. ECB euro reference exchange rates — The rate used to convert euro prices.
  3. Each tool’s own site, repository and pricing page — Every entry links to its sources and shows its check date.

Each tool’s own sources are linked from its entry. How tools are chosen and ordered, and how to correct an entry: methodology.

Cite this page

AI Pentesting Tools. “How AI pentesting tools are chosen, checked and ordered.” Updated November 8, 2026. https://aipentestingtools.com/methodology/