Show HN: An AI traffic guardrail
Building something based on 2 decades of support experience.
It's a guardrail for AI traffic where you can apply policies, redact PII and API keys, set model destinations, catch destructive actions before execution, etc.
The whole thing is built with Claude and can also be configured and troubleshot via Claude/AI tools (it acts as an MCP server).
It acts as a proxy for AI traffic. All you need is a Linux box (the Collector, where policy rules are applied) and a client app installed on the machine generating AI traffic to forward selected requests to it.
Everything is visible in the UI. Lots of customization, with user experience kept in mind throughout.
Happy to answer any questions!
Comments URL: [https://news.ycombinator.com/item?id=50330158](https://news.ycombinator.com/item?id=50330158)
Points: 1
source & further reading
kilasec.com — original article