ARTEX vs CAI:a platform in active forks against an archived framework.
CAI, short for Cybersecurity AI, is the open framework from Alias Robotics. Its GitHub repository is now archived. This page compares it with ARTEX on design, license, status and the numbers each side publishes, as of 2026-11-16.
Data as of 2026-11-16
ARTEX and CAI side by side
| Aspect | ARTEX | CAI |
|---|---|---|
| Type | A server: one Go binary with a web console over PostgreSQL | A Python framework for building security agents, used from a terminal |
| License | AGPL-3.0 | MIT parts from openai-agents-python plus proprietary additions licensed for research only |
| Open or closed | Open code. The original repo was deleted on 2026-10-08 and forks carry on | Archived. The final source tree stays readable. A paid release v1.1.5 was for customers only |
| Status | RuoJi6/ARTEX released v0.3.30 on 2026-11-14 | No more releases or fixes. Last open release 0.5.10 on PyPI in December 2025. Last push 2026-08-22 |
| What came next | Forks carry on: RuoJi6/ARTEX, BaaRai/ARTEX and sharenjun/ARTEX-main, each with its own release line | Cybersecurity Superintelligence (CSI), a commercial product from Alias Robotics |
| Architecture | A planner and 3 workers by default over an asset graph and an exploration graph, with a guard and a recording proxy | Agents built on openai-agents-python components, with a four-layer guardrail design described in a paper |
| Model support | Any Anthropic or OpenAI compatible endpoint, with a failover chain per task | OpenAI, Anthropic, DeepSeek, Ollama and others through a model setting. The alias models are not available through the archive |
| Stars | 2,090 (RuoJi6/ARTEX) | 9,862 stars and 1,553 forks |
| Public benchmark | Baidu Agent+ final: 1st overall (2026-09-03). Live range 67.83/100, 7th. TSecBench 89.78/100, 6th | A paper summarized in its README reports 41 of 45 flags and a $50,000 Neurogrid prize |
An archived repo against live forks
The biggest difference is status. The CAI README says the repository is archived, that no further releases, bug fixes, security patches or support will be provided, and that the work moved to a commercial successor, Cybersecurity Superintelligence. The source is kept as a read-only research artifact on a branch named archive, and the open releases up to 0.5.10 stay on PyPI.
ARTEX's original repository was deleted on 2026-10-08, and its last original release was v0.3.15. The code lives on in forks, and the most active line is RuoJi6/ARTEX, which released v0.3.30 on 2026-11-14. Both projects are therefore past their original maintainers, in different ways: one frozen on purpose with its code left open, the other continued by other people.
License differences
CAI's license is a mix. Its README says the project combines MIT-licensed components derived from openai-agents-python with proprietary additions that are licensed for research purposes only. GitHub's detector does not assign a single license to the repository. ARTEX is AGPL-3.0 throughout, so anyone who runs a modified version as a network service must publish the source.
If you plan to use either one in a product, read the license files in the repository itself and not only this summary.
Design and models
CAI is a framework: you build or select agents, give them tools and run them from a terminal, with guardrails against prompt injection that its authors describe in a paper. ARTEX is a finished platform with a console, a fixed planner and worker structure and a guard on every tool call.
For models, the CAI README says to set a model through its configuration and an API key for a supported provider such as OpenAI, Anthropic, DeepSeek or Ollama. ARTEX accepts any endpoint in the Anthropic or OpenAI formats and can chain several profiles.
Published numbers
The CAI README lists a research record: 18 papers, more than 30 CVE IDs and first place at several security contests. One paper it summarizes reports 41 of 45 flags and the $50,000 Neurogrid prize. The README also quotes its own framework paper as 3,600 times faster than human pentesters at 156 times lower cost. Those figures come from the authors, so treat them as claims made by the project.
ARTEX's public numbers come from the Baidu Agent+ final and TSecBench. A capture-the-flag circuit and a live attack and defense range are different tests, so this page does not rank the two tools by them.
More comparisons
CompareAntiFox vs ARTEX
AntiFox is a public fork that says it was built on ARTEX and adds report delivery. What changed, how active it is and its license.
ARTEX vs PentestGPT
A Go platform with a web console against a Python command-line agent from a USENIX Security 2024 paper.
ARTEX vs Strix
A self-hosted Go platform against an Apache-2.0 Python agent with a Docker sandbox, a cloud edition and CI support.
ARTEX vs HexStrike AI
A self-contained agent platform against an MCP server that lets your own AI client call more than 150 security tools.
Follow the ARTEX story.
New forks, releases and reports, dated and sourced.
