Hi [First name],

We’re excited to announce that Agent Vault is now in preview. Our new product lets your AI agents call the APIs and services they need without reading the credentials.

Anything an agent can read, it can leak. A credential in its environment can end up in logs, on model provider servers, or with an attacker through prompt injection.

With Agent Vault, your agent's requests go through a proxy that adds the real credential to each one on the way out. The credential stays in Infisical and never enters the agent's environment or context.

What this means for your team:

• Claude Code, Codex, and agents you build all work through the proxy with no code changes.
• If an agent gets prompt injected, it can still only reach the services in its access bundle, down to HTTP methods and paths. The proxy blocks and logs anything else.
• Each agent’s access is tied to a session that expires on a schedule you set. Every request is logged, letting you know whether it was allowed and who started the session.

You can run Agent Vault in Infisical Cloud (US or EU) or self-hosted. It works with HTTP services today, and features and pricing may change during the preview.

The setup docs help you get started.

Want to see it in action first? We’re running a live demo on October 21 at 1 PM ET. Save your spot.