Skip to main content

CDPG Data Exchange

CDPG is a Go-based platform for publishing, discovering, governing, exchanging, and auditing data across organizations. It supports API, geospatial, temporal, subscription, and file-based access, plus an optional Agentic Plane for AI agents operating under explicit delegated authority.

The platform is pre-first-release and not yet production-ready. Pages describe the intended complete architecture and label unfinished behavior explicitly. Start with Current State and Target State whenever deployment readiness matters.

Foundational attributes

  • Provider-controlled: providers and their organizations remain authoritative for resource metadata and access grants.
  • Policy-driven: authentication proves identity; a separate authorization decision governs each protected operation.
  • Standards-oriented: OIDC/OAuth 2.0, JSON-LD, NGSI-LD, OGC APIs, OpenAPI, MCP, and CloudEvents-style envelopes define interoperable boundaries.
  • Service-owned: each Go service owns its capability, data, and contracts and consumes the shared Go platform for cross-cutting behavior.
  • Default deny: unknown, invalid, expired, unavailable, or unsupported security state does not widen access.
  • Auditable: user, workload, application, agent, delegation, grant, decision, and approval context can be correlated across the platform.
  • Configuration-selected: the same images support central deployment and opt-in extensions without code branches for topology.

How it works

  1. A provider onboards resource metadata in the Catalogue and ingests data through a data service or File Exchange.
  2. A consumer discovers the resource and receives entitlement through an explicit grant, access request, or marketplace purchase.
  3. The Policy Service records the grant and publishes a durable fact; the Authorization Service updates its relationship and contextual-policy projections.
  4. The consumer calls the gateway. Authentication establishes principal identity; authorization evaluates the operation, resource, relationships, attributes, context, and delegation.
  5. The target data service verifies the carried decision, applies filters/masks/quotas and organization scoping, then returns data and emits audit signals.
  6. An enabled agent follows the same path through a governed MCP tool boundary, with additional delegation, risk, approval, and kill-switch controls.

Primary audiences

AudienceStart here
Architects and technical leadersArchitecture Overview · System Architecture
Security and risk teamsSecurity Architecture · Architecture Decisions
Platform engineers and operatorsDeployment Architecture · Monitoring
Service developersShared Go Platform · Getting Started
Data providers and consumersCore Concepts · Data Access Models
Agent developersAgentic Plane · Agent Delegation