Resources
Practical guidance for governing enterprise AI.
Architecture patterns and customer playbooks live here. Exact commands, screen paths, and release-specific behavior stay in the product documentation so there is one technical source of truth.
Customer field guides
Start with the operating decision, not a product screen
Durable guidance for designing controls before translating them into product configuration.
Designing cost-aware model routing and failover
A practical architecture for controlling model spend without turning a budget limit into an application outage.
Read field guide ArchitecturePreventing PII and secret leakage in AI traffic
A layered control pattern for prompts, model responses, and tool-call arguments that keeps policy useful without overstating what detection can guarantee.
Read field guide Security operationsIncident response for a leaked AI access key
A containment and recovery playbook that separates workload virtual keys from reusable provider credentials.
Read field guideSDK and agent patterns
Apply governance without rewriting the application
Detailed implementation patterns using established SDK and agent boundaries.
Manage OpenAI SDK access across production apps
Use workload-specific virtual keys, limits, guardrails, and auditable model access.
Secure Anthropic SDK applications
Put Claude applications behind a governed endpoint without distributing provider credentials.
Manage AI agent tool access
Separate model governance from least-privilege tool execution and approval controls.
Product documentation
Use the guide that follows the current release
Commands and UI procedures open in the GitHub-managed Agent Access Manager documentation.
Customer operations
Plan deployment, assurance, and support
Company-controlled resources for the parts around the product runtime.