AgentVault
agent-vault is an HTTP credential proxy and vault built by the Infisical team specifically for AI agents. Rather than embedding API keys, database passwords, or OAuth tokens in agent configuration or environment variables, agents request credentials at runtime through the vault proxy, which enforces access policies and never exposes raw secrets to the agent process directly.
The tool is designed to integrate with any agent harness — Claude Code, OpenClaw, Hermes, and custom setups — and acts as the credential isolation layer in an agent security architecture.
Placed in Platforms > Assess at inner position because credential management for AI agents is an unresolved security problem: agents need secrets to do real work, but current practices (env vars, config files) provide no isolation or auditability. agent-vault addresses this with a purpose-built runtime proxy rather than a bolted-on secret manager. It directly complements agent-governance-toolkit (policy enforcement) already on the radar, completing the agent security picture. Provenance from the Infisical team gives it credible integration with a battle-tested secrets platform. Worth assessing early in any production agent deployment.
