Cortex Auth launches open-source secrets vault for AI agents — lets autonomous software safely access passwords and API keys without exposing them

davideuler released Cortex Auth on GitHub as an open-source project: a specialized password and API key storage system designed for AI agents (autonomous software that makes decisions and takes actions on its own). Unlike general-purpose vaults, it uses exec-based injection — a method that feeds secrets to agents at runtime without storing them in memory or logs where they could be stolen.

Traditional secrets management forces engineers to choose between two bad options: either give agents read access to all credentials (security risk) or hardcode secrets into code (also a security risk). Cortex Auth lets teams grant each agent only the specific passwords or API keys it needs for its assigned job, checking those permissions every time the agent asks for a secret.

For companies running AI agents that handle real work — like customer service bots accessing support databases or finance agents paying vendor invoices — this removes a major blocker. Teams can now deploy agents to production without the fear that a bug or attack will leak all company secrets. Startups and enterprises already building agent systems can use it free at https://github.com/davideuler/cortex-auth.