Breaking

OpenA2A Releases HackMyAgent, Open Source Security Scanner for AI Agents and MCP Servers

October 5, 2026 at 16:16 EDT

Security
Open Source
AI Agents

An open source security toolkit designed specifically for AI agents and Model Context Protocol servers is now publicly available, offering comprehensive scanning for prompt injection, credential exposure, poisoned memory, and supply chain risks. The tool, called HackMyAgent and created by the OpenA2A organization, functions as a "Metasploit for AI agents"—providing security scanning, red-team testing, and behavioral simulation capabilities for developers building AI agent systems.

AI SECURITY · OCTOBER 2026

HackMyAgent: Open Source Scanner for AI Agent Security Threats

OpenA2A releases comprehensive security toolkit detecting prompt injection, credential exposure, poisoned memory, and supply chain risks across AI agents and MCP servers.

Coverage by the Numbers

320

Static security checks

Vulnerability categories

164

Adversarial payloads

Attack categories

Threats Detected

Agent-Specific Risks

• Prompt injection attacks
• Poisoned persistent memory
• Malicious tool exploitation
• Jailbreak attempts

Infrastructure Risks

• Exposed API credentials
• MCP server misconfigurations
• Supply chain vulnerabilities
• Dependency CVEs

Key Context

Why now: AI agents increasingly access external tools and persistent memory. Recent security analyses found 43% of MCP servers contain command injection flaws, 22% permit arbitrary file read, and 30% are vulnerable to SSRF. Agent memory poisoning persists across all future sessions, making this a critical attack vector.

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.