Open Standards & Contracts

# Open Standards & Contracts

AI Fabrix is built on a deliberate architectural principle:

Open contracts define interaction. Controlled execution enforces governance.

This separation ensures enterprise-grade inspectability, portability, and long-term architectural control.

OpenAPI as the System Contract

OpenAPI is the system-to-system contract of AI Fabrix.

It defines how external systems, applications, and services interact with the AI Fabrix Dataplane without embedding governance, credentials, or business logic into client code.

Role of OpenAPI

OpenAPI is used to:

• Expose governed dataplane capabilities
• Provide inspectable and versioned integration contracts
• Separate interface definition from execution and enforcement
• Enable portability across tools and platforms

OpenAPI describes what is available. The Dataplane decides how it is executed.

What OpenAPI Is Not

OpenAPI does not contain:

• Permission logic
• Filtering rules
• Credentials or service accounts
• Governance decisions

All enforcement remains structural inside the Dataplane and Controller.

MCP (Model Context Protocol)

MCP is the agent-to-dataplane contract in AI Fabrix.

It defines how AI agents request tools and context without direct system or data access.

MCP in the Architecture

Through MCP:

• Agents request capabilities
• The platform evaluates identity and policy
• The Dataplane executes under governance

Agents never integrate directly with systems.

MCP Responsibilities

MCP provides:

• Standardized tool definitions
• Explicit, inspectable agent interfaces
• Decoupling between agents and execution

MCP does not store data or enforce governance. Those responsibilities remain with Miso and the Dataplane.

Why AI Fabrix Avoids Proprietary SDKs

AI Fabrix deliberately avoids proprietary SDK-centric integration models.

Problems with SDK-Based Platforms

SDKs often:

• Embed security and governance assumptions
• Hide execution behavior
• Introduce vendor lock-in
• Complicate audits and exit scenarios

AI Fabrix Approach

AI Fabrix uses:

• OpenAPI for system integration
• MCP for agent interaction
• Declarative Dataplane pipelines for execution

This keeps execution governed, inspectable, and independent of client implementation choices.

Inspectability and Exit Strategy

Inspectability is a first-class architectural requirement.

Every layer exposes:

• Declarative definitions
• Explicit contracts
• Deterministic audit paths

Exit by Design

Because AI Fabrix relies on:

• Open standards
• In-tenant execution
• Declarative integration logic

Customers retain a clear exit path without losing data, architecture, or governance models.

Architecture Overview Diagram

Section Summary