LLM Cognition TypeScript MIT

perishable

perishable studies the security model for AI API access in an agent-driven world. It implements scoped, time-limited token proxies that provide secure LLM access without exposing long-lived credentials — exploring zero-trust patterns for AI infrastructure.

Primary use case

Ephemeral, scoped token proxies for LLM APIs — zero-trust credentials with model-level and token-budget scope.

How it compares

perishable is one option in a category that includes Cloudflare API Gateway, HashiCorp Vault short-lived tokens , and API-key rotation scripts. Our Compare page has the full side-by-side.