I built OpenSourceAIHub.ai to kill shadow AI prompt leaks and slash LLM bills
I’ve been working on something and wanted real feedback from folks shipping with LLMs.
As a software architect, I kept seeing the exact same painful pattern:
Developers pasting logs, API keys, internal data, or even screenshots into prompts
Zero visibility into what actually gets sent to Groq, Claude, OpenAI, etc.
Teams forced to either block AI completely… or just hope nothing leaks
Neither option felt sustainable.
So I built OpenSourceAIHub.ai — a hardened, stateless security gateway and intelligent router that sits between your code and any provider (Groq, Together.ai, DeepInfra, Mistral AI, Anthropic Claude, OpenAI, Google Gemini, xAI Grok, and open-weights models).
What it does right now:
Multi-modal DLP — Detects and redacts 28+ PII entities (including visual OCR on screenshots/images) before the prompt ever leaves your system.
Smart cost arbitrage — Routes every request to the best price + lowest latency provider in real time.
Governance controls — Per-project DLP policies, budgets, circuit breakers for runaway agents, and audit-ready logs. Fail-closed by design.
Everything stays stateless for your sensitive data: no storage or training on prompts/responses — only operational metadata (entity type, action taken, request ID, timestamp) is kept.
Built lean as a solo founder on AWS: FastAPI on ECS Fargate (isolated VPC) + Lambda + CloudFront, Next.js 15 on Amplify, DynamoDB.
Business model (transparent & high-margin):
1 million free credits on signup (enough for thousands of calls)
Managed credits: 25-30% markup on wholesale tokens
Pro BYOK: $29/mo for full sovereignty + wholesale pricing
It’s live and production-ready.
Docs: https://opensourceaihub.ai/docs
Free AI Leak Checker: https://opensourceaihub.ai/ai-leak-checker
Security details: https://opensourceaihub.ai/security
Demo video: https://youtu.be/tWrh6Gn4yJM?si=oLq453P3SFiYJzcb
