vulnerator-ai
irraz/vulnerator-ai
Pentest agent where Hermes drives a Kali container and a browser over MCP in an OODA loop
vulnerator-ai is a Docker-based autonomous pentesting setup in which Hermes Agent runs an OODA loop against a target. Its tools come from a Kali Linux container and a Playwright browser exposed over MCP.
What vulnerator-ai does
vulnerator-ai uses a deliberately minimal design with no core service, REST API, UI or gates. Four containers make up the stack: a Kali container offering tools such as nmap, whatweb, nuclei, ffuf, sqlmap and Prowler through a kali-mcp server; a playwright-mcp container with headless Chromium for the web-pentest skill; an ooda-agent container running nousresearch/hermes-agent as the loop driver and client of both servers; and a vulnerator-mcp front door.
Hermes decides what to run and launches it directly against the target through the Kali MCP. The front door exposes only the runs lifecycle on top of Hermes' API server on port 8642, so MCP clients such as Claude Code or Kiro can start, observe, steer and stop a run. It never exposes raw kali_run or chat completions. An install.sh script asks for an OpenRouter API key, generates two secret tokens and starts the stack.
Key features
- OODA loop (Observe, Orient, Decide, Act) driven by Hermes
- Kali container with a curated pentest toolset over MCP
- Headless Chromium through playwright-mcp for web testing
- MCP front door limited to starting, observing, steering and stopping runs
- install.sh generates secrets and starts the default stack
When to use it
- Running an authorized, agent-driven penetration test in a containerized lab
- Triggering and monitoring a Hermes pentest run from Claude Code or Kiro
- Studying an MCP-based design for offensive security agents
Who it is for: Security practitioners who want to test an autonomous Hermes-driven pentest workflow in containers.
How it fits with Hermes Agent
Built around Hermes Agent as the decision-making brain, running the nousresearch/hermes-agent image with its API server.
How to install vulnerator-ai
These commands are copied from the project's README. Check the repository for the latest steps before you run them.
./install.shRequirements: docker, docker compose v2, openssl and an OpenRouter API key
Note: Contains offensive-security tooling intended for use only against targets you are authorized to test.
FAQ
What is vulnerator-ai?
vulnerator-ai is an autonomous pentesting stack where Hermes Agent runs an OODA loop and uses Kali and Playwright tools over MCP. It runs as four Docker containers.
Does vulnerator-ai work with Hermes Agent?
Yes. Hermes runs in the ooda-agent container as the loop driver and keeps its OpenAI-compatible API server on port 8642.
What do I need to run vulnerator-ai?
You need docker, docker compose v2, openssl and an OpenRouter API key. The installer checks prerequisites and generates the two secret tokens.
Similar automations for Hermes Agent
All automationsRedacted showcase of the Hermes Agent setup behind Gumroad operations, with 273 skills and 78 cron loops
supermodeltools Big IronArchived prototype that ran an eight-phase SDLC in Hermes Agent, gated by a Supermodel code graph
starMagic Webnovel Writer for HermesHermes Agent skill set for writing long Chinese web novels with story contracts, memory and review
ambrosecltr FinnAI companion that lives in iMessage, with background workers, Patterns automations and a web dashboard
NicoSKOOL SEO OS AI RankingStarter kit for a human-approved SEO operating system built on Hermes Agent and a local dashboard
krusemediallc Hermes Ad AgentSkill pack that turns Hermes Agent into a Meta ads media buyer using Arcads and Meta's Ads tools
Related guides: Build agent teams with the Hermes kanban board · How to run Hermes Agent securely