Cencurity

Cencurity is a security gateway for LLM agents that protects against prompt leakage and unauthorized access. It is built for developers who need governed AI coding, connecting to existing AI agents and IDE workflows without requiring rewrites to the code.

Cencurity routes an agent's traffic through a local security gateway before it reaches the LLM provider. Requests are scanned against local security policies, only policy violations are logged, and the user's API key stays inside the IDE rather than being sent to an external server.

Cencurity shows a centralized security dashboard with every agent call, its latency, policy hits, redactions and blocks in real time. It automatically detects and blocks secrets and PII before they reach users or models, traces every agent interaction end-to-end for log analysis, proxies LLM traffic with automatic redaction, and sends webhook alerts to tools such as Slack and Jira.

Cencurity is installed as an extension from the VS Code Marketplace and supports LLM providers including OpenAI, Anthropic, Gemini, OpenRouter and any OpenAI-compatible LLM. Teams can measure a dry-run rollout before enforcing policies, and getting started means opening the project's GitHub repository and following its README quickstart.

Use cases

• Protecting an LLM coding agent inside VS Code
• Keeping an API key inside the IDE instead of an external server
• Governed AI coding for a development team
• Monitoring frontend-to-backend requests for policy breaches