Hi HN, I'm the creator of TensorWall. While building LLM-powered applications, I noticed a recurring gap: developers either give their apps a raw API key (risky) or spend weeks building custom proxies to handle rate-limiting, security, and the ever-present fear of exploding API bills. TensorWall is an open-source control plane designed to sit between your applications and your models. It gives you the visibility and guardrails needed for production. Key Features: Unified API: One endpoint for OpenAI, Anthropic, Ollama, and LM Studio Cost & Budget Control: Set hard spending limits and granular rate-limiting per app to prevent "bill shocks" Security: Prompt injection detection (PII redaction on the roadmap) Observability: Full audit logs of every request and token usage (essential for compliance)

Deploy in 60 seconds:

git clone https://github.com/datallmhub/TensorWall.git cd TensorWall && docker-compose up -d

Why Open Source? Security and financial infrastructure shouldn't be a black box. Your AI gateway should be auditable, self-hostable, and community-driven. I'm looking for brutal feedback on the architecture. What are you currently using to keep your LLM costs and security under control? GitHub: https://github.com/datallmhub/TensorWall