# CIP — Composable Integration Pipelines
The cornerstone of AI Fabrix
CIP (Composable Integration Pipelines) is the AI-native integration fabric of the Dataplane. It replaces traditional SDK-based integrations, service-account connectors, and workflow automation with a governed, declarative execution model.
What CIP Is
CIP is a declarative pipeline system that:
- Executes integrations inside the customer tenant
- Normalizes external data into a consistent metadata model
- Enforces RBAC and ABAC automatically
- Produces inspectable and auditable execution paths
- Generates OpenAPI and MCP contracts
CIP defines how data is fetched, transformed, filtered, and exposed. Authorization is enforced structurally by the platform.
Why CIP Exists
Traditional integration patterns break in AI scenarios because they:
- Drop user identity through service accounts
- Duplicate permission logic in applications
- Rely on opaque workflow engines
- Implement filtering logic inside prompts or agents
CIP removes these failure modes structurally.
AI never calls systems directly.
AI only consumes outputs produced by governed pipelines.
Declarative Pipeline Model
CIP pipelines follow a structured execution pattern:
- Fetch
- Transform
- Normalize
- Filter
- Expose
Pipelines are validated before deployment and promoted across environments (Dev → Test → Prod). Execution is deterministic and versioned.
CIP Execution Model
Declarative Pipelines (Default)
- Validated before execution
- Promoted through governed lifecycle
- Automatically generate contracts
Governed Python Execution (Exceptional)
For advanced cases where declarative logic is insufficient:
- Python execution runs inside the same Dataplane boundary
- Identity and policy enforcement remain intact
- Full audit logging is preserved
There is no governance bypass.
Contracts and Interfaces
OpenAPI Contracts (System-to-System)
CIP generates versioned OpenAPI contracts suitable for application and system consumers.
MCP Contracts (Agent-to-Dataplane)
CIP generates MCP (Model Context Protocol) tool definitions for AI agents.
MCP provides tools and context. Governance and filtering remain in the Dataplane.
Security and Governance
CIP enforces:
- Identity propagation from Miso
- RBAC and ABAC filtering
- Policy pack evaluation (egress, quotas, compliance)
- Deterministic audit logging
Every execution records:
- Identity
- Pipeline version
- Policy decisions
- Output scope
CIP Execution Flow
