morpheuslord/HackBot

What it solves

HackBot is a teaching demo of an agentic AI system designed for cybersecurity research and defense. It moves beyond simple LLM wrappers by providing a multi-agent orchestration system that can perform complex tasks like analyzing security reports, researching vulnerabilities, and reviewing source code or logs without generating exploits or malware.

How it works

The system uses a multi-agent architecture managed by an orchestrator agent (via pydantic-ai) that plans tasks and spawns specialist agents in parallel. These specialists include researchers, SARIF analysts, document analysts, and tutors.

Key technical components include:

• Model Context Protocol (MCP): Tools are hosted in separate MCP server processes, allowing agents to act as clients and access tools like web search, CWE lookups, and file loaders.
• Agent Communication Protocol (ACP): Agents are published via an ACP server, making them discoverable and callable by other ACP clients or the orchestrator.
• Structured Output: Specialists return typed Pydantic models rather than free text to ensure consistency.
• TUI Interface: A Rich-based terminal application provides a dashboard, streaming chat, a live activity tree for inspecting agent steps, and interactive windows for triage and file review.

Who it’s for

Cybersecurity students and researchers who want to learn how to build and use agentic AI systems for defensive security analysis.

Highlights

• Parallel Agent Execution: The orchestrator can spin up multiple researcher agents simultaneously to gather information.
• Flexible Input: Supports any text file or folder (logs, source code, config files) for analysis via a dedicated document analyst.
• SARIF Integration: Specialized triage for SARIF reports with an interactive review window for accepting or rejecting findings.
• Extensible Tooling: Supports adding extra MCP servers via a standard JSON configuration file.
• Safety-First: System prompts explicitly refuse the generation of exploits or malware.