Runs configurable MCP and network decoys to detect prompt injection and observe malicious agent behavior.
Copy the AI prompt to install this server into Claude Code, Cursor, or another agent β or use 1-click editor setup below.
One-click editor setup isnβt available for this listing yet β we donβt have a confirmed install command, and weβd rather show nothing than point your editor at the wrong package or host. Follow the projectβs own setup instructions, linked above.
Inspect callable tools, capabilities, and parameters exposed to AI agents by Beelzebub.
The mariocandela/beelzebub MCP server is a deception runtime for creating decoy services intended to attract and record suspicious behavior. Its supported protocols include MCP, SSH, HTTP, TCP, and TELNET. The MCP service is relevant to agent-security testing: an agent can be given tools or interactions outside its normal task profile, allowing operators to observe whether it follows malicious instructions or attempts prompt injection.
Service behavior can be described with YAML rather than implemented entirely in code. Regular-expression command matching determines which configured response applies. Optional LLM support from OpenAI or Ollama can produce context-sensitive responses during an interaction.
The runtime loads a core configuration file and a directory of service definitions, then starts the enabled deception services. The run command accepts separate paths for these inputs and applies a configurable per-service memory limit. The validate command parses and checks the configuration without starting listeners, which makes it suitable for CI validation.
Extension points are provided through the Go SDK in pkg/plugin. Developers can implement command-response plugins, HTTP handlers, or binary TCP wire plugins. Plugins are registered through Go initialization and can be installed from GitHub, then included in a local or Docker build. TCP services can also define message framing and explicitly order wire plugins for stateful binary exchanges.
The repository provides an installer with local and Docker modes. Local development can also use make start; Docker builds and runs through make docker. A Helm chart is included for Kubernetes deployments. The documented default configuration locations are ./configurations/beelzebub.yaml for core settings and ./configurations/services/ for service definitions.
The CLI binary provides run, validate, plugin, and version commands. plugin install fetches an external plugin from GitHub and can append it to the plugin configuration. The projectβs local build path requires Go, while the Docker path requires Docker; the excerpt does not specify an MCP client installation command or client-specific configuration.
The mariocandela/beelzebub MCP server supports:
This project is a deception and observation runtime, not a general-purpose MCP tool provider. Its usefulness depends on the decoy services and configuration supplied by the operator. The excerpt does not document a hosted endpoint, a required API key, supported desktop clients, or environment-variable configuration. OpenAI and Ollama are listed as LLM options, but their individual setup requirements are not specified here.
The MCP protocol is one part of a broader multi-protocol runtime, so deployment may involve listeners and service definitions beyond an MCP-only test. External plugins also introduce a build and installation step before they become part of a local or Docker runtime.
Factual signals from GitHub, npm, and our automated checks β not a rating.
No reviews yet β be the first to share how this listing worked for you.
Showcase your server listing on GitHub or your project documentation. Embed this dynamic SVG badge to highlight official listing status and live engagement.
[](https://allmcps.com/mcp/mariocandela-beelzebub)<a href="https://allmcps.com/mcp/mariocandela-beelzebub"><img src="https://allmcps.com/api/badge/mariocandela-beelzebub?style=directory" alt="Beelzebub on AllMCPs" /></a>