Compatibility and deployment requirements.

Understand where AgentFleetGateway fits today and what should be confirmed for your environment.

Available today#

• Kubernetes-native agent registration, reconciliation, routing, scheduling, policy, approvals, audit, and Console workflows.
• OpenAI, Anthropic, Bedrock, and approved provider-compatible model endpoints.
• Optional deterministic Guard inspection for configured model calls.
• Optional protected model access with workload identity, controlled egress, server-side credentials, and agent suspension.
• Optional Istio configuration, Prometheus metrics, ClickHouse audit views, Grafana dashboards, and expert CLI workflows.

Confirm for your deployment#

• Your Kubernetes platform, CNI, ingress controller, storage classes, and service-mesh configuration.
• Your identity provider, tenant claims, role mappings, secrets platform, and data-residency requirements.
• Your model-provider credentials, quotas, regional endpoints, and cost controls.
• Your MCP authentication flows, A2A trust relationships, and external-agent endpoints.
• Your availability, backup, recovery, retention, and incident-management objectives.

Current boundaries#

• Global active-active control-plane replication and placement-aware multicloud scheduling are not included.
• Managed Kubernetes and service-mesh combinations are supported after validation with the customer environment.
• Guard adds configured inspection controls; it does not guarantee factual accuracy or prevent every unsafe model response.
• Third-party MCP servers, A2A peers, and external agents remain subject to their own security and availability characteristics.