Use visual construction where it helps. Use code where it matters.
Benefit: Use the development approach appropriate to each requirement without creating a separate architecture for everything the platform did not build natively.
Enterprise applications eventually encounter requirements that generic components cannot anticipate.
Teams may need proprietary business logic, specialist algorithms, high-performance processing, existing enterprise services, industry-specific APIs, custom models, specialist data infrastructure, proprietary user interfaces, and provider-specific capabilities.
Frozion does not require those requirements to be rebuilt as visual configuration. Engineering teams can extend the application through defined capability boundaries while preserving the application semantics surrounding the extension.
Preserve source-to-runtime traceability
A deployed extension is more useful to engineering, security, and audit teams when they can identify which implementation or registered external capability participated in the running application.
For capabilities built and deployed by the organization, the extension lifecycle can preserve relationships such as source revision, build artifact, capability version, published application, Execution Instance, and evidence.
Benefit: Connect engineering and capability change history to the application behavior that actually ran.

Put every extension behind an explicit capability boundary
Benefit: Give custom code a governed place in the application rather than treating it as an opaque escape hatch.
Inputs
What information enters the capability.
Outputs
What the capability returns.
Schemas
The shape and meaning of the data crossing the boundary.
Identity and authority
Which subject, workload, tenant, purpose, and delegated authority apply.
Policy
Which conditions determine whether the capability is eligible to run.
Side-effect behavior
Whether the capability reads, calculates, proposes, persists, or changes an external system.
Versions
Which implementation and contract version belong to the published application.
Evidence
What execution information, receipts, or outputs remain afterward.
Bring Custom And External Capabilities Into Governed Execution
Most extensions exist because a requirement crosses into execution: an existing service, proprietary logic, a specialist algorithm, a particular model, or a system of record. Each of these enters through the same capability boundary.
Connect existing services through APIs
Expose business systems, pricing engines, risk services, identity services, document processors, payment systems, CRM or ERP functions, internal microservices, and third-party services through explicit API contracts and mappings.
Benefit: Reuse existing enterprise services without allowing every API boundary to become another place where application meaning is lost.

Build custom execution capabilities
Create specialized capabilities for deterministic calculations, algorithms, transformations, parsing and extraction, validation, optimization, classification, proprietary decision logic, and high-performance computation.
Use high-performance kernels where the workload needs them
Performance-sensitive application steps can use specialized deterministic kernels while still exposing typed inputs and outputs and participating in policy, versions, routing, and evidence.

Bring your own models and providers
A governed application should be able to use specialized AI capabilities without making one model provider the application architecture.
A model or provider adapter can describe supported task, input and output contract, model or capability version, provider, deployment route, data eligibility, region, policy, quality profile, routing eligibility, and evidence requirements.
Benefit: Adopt new models and providers without redesigning the application around each provider's API.
Explore Model & Capability RoutingKeep provider-specific capabilities first-class
Provider portability should not require reducing every cloud or AI provider to the lowest common denominator. Cloud agnostic does not mean cloud generic.
Benefit: Take advantage of differentiated cloud and provider capabilities without making implementation details own the complete application.


Preserve source permissions when data enters the application
Connectors bring external data and capabilities into the application while retaining tenant, source identity, source permissions, taxonomy, classification, schema, version, sync state, and provenance.
Benefit: Add enterprise sources while preserving the scope and provenance needed to use their information correctly.
Package reusable logic as subgraphs
Reusable subgraphs let teams define governed patterns once and include them inside larger Execution Graphs with explicit contracts, policy, versions, dependencies, and evidence requirements.
Extend policy with typed functions
Specialized policy functions can receive defined facts, return defined results, carry versions, execute under known authority, produce evidence, and participate in testing and enforcement.
Use the same deployment choices for custom capabilities
Extensions should participate in the workload-specific deployment model. Deployment rules can remain part of the application's capability and policy model.
Benefit: Place custom capabilities where they belong without creating a separate deployment architecture for extensions.
Explore Private & Hybrid AISee extensibility in one business application
The core application may already provide the portal, operational records, Context Assembly, Execution Graph, policy, approval, and evidence. Engineering can extend that application where specialized capabilities are required.
Custom account-validation service
Invoke an existing enterprise service to validate account ownership or format.
Specialist risk capability
Run proprietary deterministic or machine-learning logic against the proposed account change.
External identity provider
Use existing enterprise or supplier identity infrastructure.
Document connector
Retrieve evidence from an existing repository.
Provider-specific AI service
Use an eligible cloud or private AI capability when policy allows.
Custom reviewer component
Present a specialized banking or risk review experience.
Explore Developer Extensibility
APIs And Custom Capabilities
Connect services, algorithms, and deterministic capabilities through governed contracts.
Explore Capability ContractsBring Your Own Models
Register private, hosted, provider-specific, and future model capabilities.
Explore Model RoutingReusable Subgraphs
Package governed workflow and application patterns for reuse.
Explore Execution Graph