AI-Infra-Guard is Tencent’s open-source, full-stack AI red teaming platform for securing AI systems and infrastructure. It scans agents, skills, MCP servers and AI infrastructure components for vulnerabilities, and evaluates large language models against jailbreak attacks. The project has earned about 6,900 stars on GitHub, is Apache 2.0 licensed, and is under very active development in 2026.
As companies deploy agents, MCP servers and model-serving stacks, the attack surface grows faster than traditional security tooling can follow. AI-Infra-Guard was built by Tencent’s security team to cover that whole stack in one platform. Its Agent Scan, Skills Scan and MCP Scan modules inspect the components an agentic system is built from, flagging risky tools, prompt-injection paths and unsafe configurations; the AI Infra scan checks deployed frameworks and services against a fingerprint database of known AI components and their CVEs; and an LLM jailbreak evaluation module stress-tests models with adversarial prompt suites. Findings are presented through a web console with reports a security team can act on, and scans can be run on demand or integrated into deployment checks, making it a practical scanner for anyone shipping AI infrastructure in production.
Free. AI-Infra-Guard is open source under the Apache 2.0 license and self-hosted, with no paid tier.
Security engineers, platform teams and AI developers who deploy agents, MCP servers or model infrastructure and need to find vulnerabilities and jailbreak weaknesses before attackers do.