Agent: Claude Code, CursorLLM: Claude 3.5, GPT-4#credential-management#ai-agents#security#proxy#prompt-injection-defense
An open-source credential broker that sits between AI agents and APIs they call, preventing credential exfiltration through prompt injection. Agent Vault brokers access to services like LLM providers and GitHub without agents holding real credentials, using transparent HTTP proxying and pluggable credential stores.