Beelzebub

Beelzebub is an open source deception runtime that deploys decoy services for both infrastructure and AI agent attack surfaces. It emulates SSH, HTTP, TCP, TELNET and MCP services, lets operators define realistic decoys in YAML, collects detailed events whenever an attacker interacts with one, and can be extended with Go plugins. Because a decoy has no legitimate purpose, any interaction with it is a strong signal of malicious activity. The documentation addresses three audiences. Security operators install, configure, expose, monitor and retire decoys, with deployment guides for Docker, Docker Compose, Kubernetes and Helm and optional integrations with RabbitMQ, Prometheus and LLM providers. AI security teams deploy MCP tripwires and LLM powered interactions to measure how often agents bypass their guardrails. Developers use the public plugin SDK to add protocols. A production safety checklist stresses written authorization, isolation from production credentials, restricted outbound traffic and synthetic secrets, and the project states that the runtime is a detection component, not a security boundary. The runtime is free and open source. The company behind it also sells a commercial active defence platform with three products: Arcangelo for internal and external attack surface management and agentic penetration testing, Beelzebub Platform for managed deception with canary credentials and AI led investigation, and Caronte for agentic malware and infrastructure analysis using the open source Azazel sandbox. Commercial pricing is not published; buyers start by scoping a proof of value with the vendor. The documentation tracks the main branch of the repository, so it can describe features newer than the latest tagged release.