Eight authentication mechanisms, multi-tenant by construction, audit by default.
A unified dynamic dispatcher routing to eight authentication mechanisms — HTTP Basic, Digest, Form + 2FA, JWT, OAuth 2.0, OpenID Connect, SAML 2.0 and mTLS. Multi-tenancy enforced through scoped per-request context at the runtime level. Row and column security injected into every query. Audit is a property of every write, not a feature added on top.
Eight mechanisms, one perimeter
HTTP Basic, Digest, Form + 2FA (TOTP / Email / SMS), JWT, OAuth 2.0, OpenID Connect, SAML 2.0, mTLS — unified under a single dynamic dispatcher. Mix LDAP, SSO and certificates inside a single installation.
Multi-tenant by construction
The runtime binds the authenticated principal and the target database to the request via scoped values (JEP 446). Applications cannot bypass tenant scoping; the runtime enforces it.
Row and column security injected by the compiler
The query compiler — not the application, not any AI agent that authored the query — injects tenant-scoping WHERE clauses, row-level expressions and column-level visibility rules into every generated SQL statement. Security is a property of the compilation, not a discipline the application has to remember.
Audit is a property of every write
Every transaction, master-data change and approval is recorded with timestamp, user identity, source document and prior-value history. The audit trail is the record, not a log of it.
Eight authentication mechanisms
Dispatched per request by a single dynamic authenticator. Same installation can authenticate users from LDAP, OAuth, SAML and client certificates concurrently.
HTTP Basic
Base64-encoded credentials over TLS. For service-to-service integrations where token lifecycle is unnecessary.
HTTP Digest
Challenge-response authentication; credentials never transmitted in plaintext.
Form + 2FA
Browser-form login with TOTP (RFC 6238), email OTP, SMS OTP or CAPTCHA. The default for interactive users.
JWT Bearer
Stateless signed-token validation. Signature, claims and expiry verified per request. Per-user revocation salt for instant invalidation.
OAuth 2.0
Authorisation-code flow, token introspection, refresh tokens. Integrates with enterprise identity providers.
OpenID Connect
OIDC discovery, ID-token validation, userinfo endpoint, claim mapping to platform principals.
SAML 2.0
SP-initiated SSO, assertion parsing, signature verification. The standard for enterprise federation.
mTLS
X.509 client certificates with DN-to-principal mapping. For service-to-service trust in regulated environments.
Multi-tenancy and access control
Tenant isolation is a property of the runtime, not a discipline applied by application developers.
Scoped per-request context
Every request runs inside a scoped context that binds the authenticated principal, the target database and the tenant identity. Scoped values (JEP 446) make it impossible for application code to read out-of-scope context.
Per-tenant database routing
The bound database in the request context determines which JDBC pool, which engine and which replica the request hits. Tenants share infrastructure; they never share data.
Row-level security — injected by the compiler
Per-object SQL expressions injected by the platform's query compiler into every SELECT. Roles, departments and ownership rules are enforced at the data tier — not by application code, not by any AI agent that authored the query, not by any layer that could be bypassed. The security expression is a property of the user's runtime context; the compiler resolves it into the WHERE clause every time.
Column-level security — injected by the compiler
Field-level visibility rules per role, applied during query compilation. Sensitive columns are not just hidden from the UI — they are removed from the generated SELECT for users not entitled to read them. The data never leaves the database for users without entitlement.
RBAC with explicit deny-by-default
Role enforcement applied at endpoint level, before handler execution. In-handler role and policy checks for data-dependent rules. No way to bypass.
Audit by default
Every write is recorded with full context — user, timestamp, IP, mechanism, source document, prior-value history. A separate audit channel records authentication events independently of the data-write log.
Defence in depth, applied per layer
Security on the platform is not a single feature. It is a chain of guarantees applied at each layer the request passes through — TLS at the network edge, IP rules at the perimeter, OWASP filter at the HTTP layer, the eight-mechanism authenticator at the application edge, scoped tenant context in the runtime, row and column security at the SQL layer, mandatory audit at the persistence layer. Each layer makes assumptions about the layer below; none of them trust the application above.
The result is a posture where misbehaving application logic cannot leak data, cannot impersonate another tenant, cannot bypass rate limits and cannot escape the audit trail. The application's job is to express business logic; the platform's job is to enforce that the logic operates inside the rules.
The same architectural property covers AI agents. Agents on the platform generate XDBL (the platform's XSD-described query grammar), not raw SQL; the compiler emits engine-native SQL with row-level and column-level security injected per the user's runtime context. The agent does not connect to the database; the compilation step sits between the agent and the data. See platform/ai-mcp.html for the full chain.