AI is writing real production code — but security still happens after generation. Cencurity Engine (CAST) enforces security during generation, in real time. It runs inline between your IDE/agent and the LLM and: - blocks dangerous code (eval, subprocess) - redacts secrets (API keys, credentials) - enforces policies in the stream Works across OpenAI, Claude, Gemini, and OpenAI-compatible models (xAI, DeepSeek, LLaMA). No plugin required. 100% open source.
Hey everyone 👋
We built Cencurity Engine (CAST) because AI is now writing real production code — but security workflows are still post-generation.
We wanted to move security into the generation process itself.
So instead of scanning code after it's written, Cencurity enforces policies inline while the model is generating.
Quick demo:
- blocks dangerous code (eval, subprocess)
- redacts secrets (API keys, credentials)
- works with streaming LLMs
Supports OpenAI, Claude, Gemini + OpenAI-compatible models.
Would love direct and honest feedback 🙏
Cencurity