Infisical/agent-vault

What it solves

Agent Vault prevents "credential exfiltration," where AI agents are tricked via prompt injection into leaking sensitive API keys or secrets. Instead of giving agents direct access to credentials, it acts as a secure broker that manages secrets and injects them into requests on the fly.

How it works

It operates as a MITM (Man-in-the-Middle) HTTP proxy. You store your real credentials (e.g., Anthropic or GitHub keys) in the vault and configure the agent to route all outbound traffic through the Agent Vault proxy using HTTPS_PROXY. When the agent makes a request using a dummy placeholder (like __anthropic_api_key__), Agent Vault intercepts the request, substitutes the dummy value with the actual secret, and forwards it to the target API.

Who it’s for

Developers building or deploying AI agents—such as remote coding agents, all-purpose agents, or custom agents in ephemeral sandboxes—who need to ensure that agents cannot access or leak the raw credentials they use to interact with external services.

Highlights

• Credential Brokering: Substitutes dummy values in headers with real credentials without the agent ever seeing the secret.
• Transparent Integration: Works with existing tools (MCP, CLI, SDKs) by bootstrapping the environment with proxy settings.
• Egress Filtering: Allows operators to control exactly which agents can access specific services and API endpoints.
• Pluggable Stores: Supports local encrypted storage or external secrets managers like Infisical for dynamic secrets.
• Request Logging: Provides visibility into authenticated traffic for monitoring and diagnosing agent behavior.