Agent Armor vs QSAG-Core

Agent Armor is a lightweight governance layer for AI agents, written in Rust and designed to intercept every agent action before execution. It sits in front of LangChain, CrewAI, AutoGen, or Claude Code and runs each proposed action through an 8-stage decision pipeline: intent classification, credential leak scanning, rate limiting, resource scoping, behavioral fingerprinting, semantic deduplication, human-review escalation, and final allow/block. The project is MCP-aware and can intercept tool calls at the protocol level, which means it works regardless of which agent framework you're using. Actions that pass all 8 layers execute normally; those that fail can be automatically blocked, held for human review, or rewritten to a safer equivalent. A live dashboard shows agent activity, pending reviews, and anomaly alerts. Version 0.3.0 arrived as a Show HN today and hit the front page. The author, Edoardo Bambini, built it after a production incident where a coding agent attempted to overwrite git history on the main branch. The timing is good — as more teams ship agents to production, "what guardrails do I put between the agent and the real world?" is an increasingly urgent question.

QSAG-Core is a Python security scanner specifically designed for the OWASP Top 10 for Agentic Applications 2026 threat model. It provides three core detection capabilities: MCP tool poisoning (26 malicious patterns across 7 categories), prompt injection (28+ attack patterns including goal hijacking, jailbreak attempts, and memory poisoning), and ghost agent detection for unauthorized API key usage. It runs as pure pattern matching — no ML, no cloud dependency — and can be integrated as a pre-execution guard in any Python-based agent pipeline. Released April 10, 2026 by the Neoxyber team, QSAG-Core fills a real operational gap as MCP-based agent deployments proliferate. While Microsoft's Agent Governance Toolkit addresses similar territory, it's heavyweight and enterprise-focused. QSAG-Core is a pip install and a few lines of code — the security-focused indie alternative that fits into a CI/CD pipeline or an existing agent framework without an enterprise contract. The threat model it addresses is timely. As MCP becomes the de facto standard for tool-calling in AI agents, malicious MCP servers and prompt injection via tool outputs are becoming documented attack vectors. Having a lightweight, open-source scanner that specifically targets these patterns is exactly what the community has been building toward. MIT licensed, 24 commits in its first day.