Agent Access Manager documentation
Agent Access Manager is a self-hosted gateway and control plane for governed AI model access. Your applications call one OpenAI-compatible endpoint with revocable virtual keys; the gateway routes across vendors, enforces budgets, rate limits, and guardrails, and records every call in a durable audit trail. These docs cover everything from the first install to the full API surface.
The shortest path to a working setup: connect a provider, mint your first virtual key, and make your first governed request in about five minutes.
Start here →InstallationRun the installer on a single host, pick an edition, and get the full stack — gateway, dashboard, database, and SSO — up with one command.
Install the product →Getting started
What the product is, the fastest path to a first governed request, requirements, installation, and the architecture behind it all.
Core concepts
How the product is organized and configured: tenancy and scopes, roles, keys, the model catalog, routing, guardrails, limits, and SSO.
Guides
End-to-end walkthroughs for real operational tasks, each with the exact steps and a way to verify the result.
API reference
Every endpoint on both planes: the OpenAI-compatible Gateway API your applications call, and the Admin API behind the dashboard.
Integrations
Point your existing SDKs, frameworks, model servers, and identity provider at the gateway — usually by changing only a base URL and a key.