# Introduction What Mandavo is, the problem it solves, and how the fabric fits your stack. ## What Mandavo is Mandavo is the identity and access fabric for AI agents. Every agent gets a distinct cryptographic identity, short-lived credentials scoped to the task in front of it, and a named human sponsor who is accountable for what it does. It sits upstream of your agent tooling as the neutral issuance-and-scoping layer. It is cross-vendor by design and sponsor-bound by default, so the same controls apply whether an agent runs on your models, a partner's, or an off-the-shelf framework. ## The problem it solves Most agents today run on borrowed logins and shared API keys that never expire. When one is compromised, there is no clean way to say which agent acted, who was responsible, or how to contain it without rotating secrets across your whole stack. Mandavo replaces standing keys with credentials that expire before they can leak, attributes every action to one identity and one sponsor, and lets you invalidate a compromised agent everywhere at once — in seconds, on the record. ## The four modules Agent Identity Directory — the canonical registry of who exists, who sponsors each agent, and what scopes each may hold. Issuance & Rotation Engine — short-lived, task-scoped credentials provisioned on demand and rotated automatically. Policy Graph — a machine-readable statement of who may do what, under which conditions, evaluated at execution time. Audit Log — an attestation-ready record of every decision, from which SOC 2 and ISO 27001 evidence generates itself. ## Where to go next Run the Quickstart to register your first agent, issue a scoped credential, and revoke it in about ten minutes. Read Core concepts for the vocabulary — identities, sponsors, scopes, the policy graph, and the audit log — used throughout these docs.