Launching today
Overrule
Control every LLM call in real time
3 followers
Control every LLM call in real time
3 followers
The open-source governance layer for production AI. Intercept every LLM call, enforce policies locally in sub-millisecond, stream proof to a cloud audit trail. No proxy, no vendor lock-in, $0 to start.
Hey — Rashid here, I built Overrule.
Quick backstory: I've been building backend systems for 4+ years. Over the past year, I saw multiple teams ship LLM features to production with basically no guardrails.
No PII filtering. No prompt injection protection. No proper logs.
That didn't sit right — especially with things like the EU AI Act coming in.
So I built Overrule.
It's a simple Python SDK that sits between your app and any LLM. It checks every request/response for PII, prompt injection, toxicity — and blocks anything that shouldn't go through.
It also logs everything to a dashboard so you're not blind when something goes wrong (or when someone asks for compliance).
It's open-source (MIT), has a free tier, and takes 5 minutes to set up.
Would genuinely love feedback — especially what policies you'd want beyond PII / injection / toxicity. That's exactly what I'm building next.
Local policy enforcement actually feels instant, way faster than the proxy setups I have wrestled with before. Streaming the audit trail to the cloud without leaking payloads is the part that sold me.
@kurklu92363 Thanks! That's exactly the tradeoff we obsessed over — governance shouldn't tax your latency or route data through someone else's infra. Policies run in-process, sub-ms. Only structured metadata hits the cloud — never your payloads.
Appreciate you trying it out. Hit us up on GitHub if you need anything.
love that this runs locally with no proxy overhead. one thing that would help us adopt it faster is built-in policy templates for common frameworks like SOC 2 and HIPAA, so we don't have to author every guardrail from scratch on day one.