Authentication
Virtual MCP implements a two-boundary authentication model that separates client authentication from backend authentication, giving you centralized control over access while supporting diverse backend requirements.
Two-boundary authentication model
Boundary 1 (Incoming): Clients authenticate to Virtual MCP using the mechanism defined in the MCP specification. This is your organization's identity layer.
Boundary 2 (Outgoing): Virtual MCP obtains appropriate credentials for each backend. Each backend API receives a token or credential scoped to its requirements.
Incoming authentication
Configure how clients authenticate to Virtual MCP.
Anonymous (development only)
No authentication required. Use only for local development.
spec:
incomingAuth:
type: anonymous
OIDC authentication
Validate tokens from an external identity provider.
Kubernetes service account tokens
Authenticate using Kubernetes service account tokens for in-cluster clients.
Outgoing authentication
Configure how Virtual MCP authenticates to backend MCP servers. Outgoing
authentication is configured through MCPExternalAuthConfig resources.
Discovery mode (recommended)
When using discovery mode, Virtual MCP checks each backend MCPServer's
externalAuthConfigRef to determine how to authenticate. If a backend has no
auth config, Virtual MCP uses the specified default.
spec:
outgoingAuth:
source: discovered
default:
type: discovered
External auth config reference
Reference an MCPExternalAuthConfig resource for specific backends:
spec:
outgoingAuth:
source: inline
backends:
github:
type: external_auth_config_ref
externalAuthConfigRef:
name: github-auth
The MCPExternalAuthConfig resource defines the actual authentication strategy:
- Token exchange: Exchange the client's token for a backend-specific token using RFC 8693
- Header injection: Inject a static header value (such as an API key)
Authorization with Cedar
Add fine-grained authorization using Cedar policies:
See Cedar policies for policy syntax.
Token caching
Virtual MCP caches exchanged tokens to reduce latency and load on identity providers:
spec:
tokenCache:
provider: memory
memory:
maxEntries: 1000
ttlOffset: 5m # Refresh 5 minutes before expiry