Self-hosted multi-agent system that autonomously plans and runs penetration tests using tools like nmap, Metasploit and sqlmap in Docker.
Scans git repos, files, and stdin for hardcoded secrets such as passwords, API keys, and tokens.
Zero-trust runtime that sandboxes and snapshots large numbers of untrusted agent workloads on standard Kubernetes.
Autonomous API and web-app penetration testing tool that runs a swarm of AI agents to chain exploits and prove vulnerabilities.
Distributed security asset discovery platform covering subdomain enumeration, port scanning, vulnerability scanning, and page monitoring.
Extensible HTTP/1-2-3 web server written in Go that obtains and renews TLS certificates automatically by default.
Agent Mesh is a private network that lets AI agents publish and discover tools and models, with verifiable node identities.
Open-source, community-governed fork of HashiCorp Vault for storing, encrypting, and distributing secrets, certificates, and keys.
Scans repositories, filesystems, and chats for leaked credentials, then verifies whether each one is still live.
Reverse proxy that adds OAuth2 and OIDC authentication in front of web applications, supporting many identity providers.