Superagent

Secures AI agents by blocking prompt injections, malicious tool calls, and data leaks in runtime. An AI security service for your repos, web apps, and agents. It finds vulnerabilities and fixes them for you. Free for public repos. SuperagentLM is the state-of-the-art safety model that powers Superagent, analyzing prompts and responses in real time to block threats with sub-50ms latency. Configure it as a proxy in superagent.yaml with your OpenAI API base, then route calls through Superagent for automatic filtering. Self-hosting provides full data control and deploys on-premise via Docker for enterprise needs. It stops prompt injections, data leaks, malicious tool calls, and backdoors in AI agent workflows. Insert checks into GitHub Actions or similar to scan code for vulnerabilities before shipping. Policies are highly customizable via YAML configs, allowing rules for specific models, providers, and threat types.