I work at Securetron, and this is something we built for people running AI agents on their own infrastructure.
AI-FW sits between your AI agents and the LLM backend. It inspects prompts and responses, enforces policy, handles routing, and keeps an audit trail, so the agent traffic is governed at the gateway instead of inside each agent.
It also does semantic-gated prompt compression to save tokens. It only rewrites a prompt when the meaning stays close (cosine similarity 0.92 or higher) and the saving is over 5 percent, so you spend fewer tokens without changing what the prompt means.
It is self-hosted. The free Community Edition covers 10,000,000,000 tokens per rolling 30 days, no credit card required.
If you self-host agents, I would like to hear how you are governing their traffic today, and what is missing. Happy to answer technical questions here.
here’s the neat thing: I don’t.
_ _ _ _| || |_ | | |_ __ _|_ __ _____ _____ _ __ ___| | ___ _ __ _| || |_| '_ \ / _ \ \ / / _ \ '__/ __| |/ _ \| '_ \ |_ __ _| | | | __/\ V / __/ | \__ \ | (_) | |_) | |_||_| |_| |_|\___| \_/ \___|_| |___/_|\___/| .__/ | | |_|