Practical writing on credential management, zero-trust architecture, and building AI agents that security teams can approve.
Long-lived credentials in agent contexts create an exposure surface that's fundamentally different from human authentication. Here's the threat model.
AWS STS got it right for cloud resources. Here's how to apply the same just-in-time model to every credential an AI agent might need.
A practical guide to every method available today — environment variables, tool injection, vault references, and just-in-time nonces — with a security analysis of each.
Most teams underestimate how many places their agent's credentials can leak. We mapped it out — from system prompts to cloud logs to memory stores.
Zero-trust says "never trust, always verify." AI agents test this model in new ways. Here's how to apply zero-trust thinking to agent identity and credential access.
10,000 nonces/month free. No credit card required.
Generate your first nonce — free