Wales Agent Infrastructure
Canonical AI Endpoints for AI Agent Discovery & Interaction in Wales
AIAgent.Wales × Wales.Team × Wales.1
Overview & National Architecture
The Canonical Endpoints Dual Stack proposes a national-scale infrastructure through which humans, organisations, software systems, and autonomous AI agents can discover, identify, evaluate, and interact with AI agents operating within or serving Wales.
Three Complementary Namespaces
- AIAgent.Wales: The human-facing discovery, trust, and interaction interface.
- Wales.Team: Identifies and describes teams, assemblies, and orchestrations of AI agents.
- Wales.1: Provides the canonical logical identity of an individual AI agent.
Dual Stack Paradigm
| Human Stack (AIAgent.Wales) | Machine / Agent Stack (Wales.1 & Wales.Team) |
|---|---|
| Human-readable agent profiles | Machine-readable manifests |
| Search and conversational discovery | Capability-based discovery APIs |
| Welsh and English interfaces | Structured bilingual metadata |
| Trust and provenance presentation | Verifiable credentials & attestations |
| Human interaction & service journeys | Agent-to-agent protocol invocation |
Strategic Core Proposition
“The Web gave organisations canonical locations. The Canonical Endpoints Dual Stack gives AI agents canonical identities, teams, and trusted means of discovery.”
Architectural Principles
The infrastructure establishes an identity, discovery, trust, and interaction layer for an emerging AI-agent economy while remaining independent of any single model provider, cloud platform, or framework.
1. Execution Independence
Canonical agent identity remains stable even when physical implementations change. A canonical identity maps cleanly to variable execution endpoints:
↓
Live Implementation: https://agents.example.org/v3/funding
2. Protocol Neutrality & Federation
Canonical identity sits above specific communication standards like MCP (Model Context Protocol) and A2A (Agent-to-Agent). The framework supports federated operations across Welsh Government, local authorities, NHS bodies, universities, businesses, and third-sector participants.
3. First-Class Welsh & English Representation
Bilingual support is represented at the schema and metadata level rather than added as a translation layer post-execution, directly aligning with the Welsh Government’s AI Plan.
Four Foundational Functions
Discovery: What can it do & where?
Trust: Who operates & verifies it?
Interaction: How to communicate?
Namespace Architecture & Identity
The architecture implements a pragmatic separation between public DNS hostnames and logical identifiers to allow immediate production deployment.
| Namespace | Role & Logical Identifier Standard | HTTPS Resolution Mapping (v0.1) |
|---|---|---|
| AIAgent.Wales | Human discovery, UI profiles, conversational search | https://AIAgent.Wales/business/funding |
| Wales.Team | Canonical identity for multi-agent teams | wales.team/business/start-a-business |
| Wales.1 | Canonical logical identity for individual agents | https://id.aiagent.wales/business/funding |
Wales.Team Composition Types
Teams represent dynamic or fixed capabilities assembled from multiple agents (e.g., funding, planning, skills, and premises agents coordinated together):
- Static Teams: Deliberately specified and strictly governed member agents.
- Dynamic Teams: Members selected through capability discovery at execution time.
- Hybrid Teams: Fixed core agents combined with dynamically discovered domain specialists.
Canonical Identity & Manifest Specification
Every registered agent possesses an immutable Agent ID resolving to a canonical manifest expose six core logical resources: Identity, Capabilities, Interaction, Trust, Policy, and Status.
Minimum Agent Manifest Schema (v0.1)
“spec”: “wales-agent-manifest/0.1”,
“id”: “wales.1:business/funding”,
“canonical”: “https://id.aiagent.wales/business/funding”,
“name”: {
“cy”: “Asiant Cyllid Busnes”,
“en”: “Business Funding Agent”
},
“operator”: {
“id”: “org:example”,
“name”: “Example Organisation”
},
“capabilities”: [
“business.finance.grants”,
“business.finance.support”
],
“languages”: [“cy”, “en”],
“coverage”: { “jurisdiction”: [“Wales”] },
“interaction”: [
{ “type”: “a2a”, “endpoint”: “https://example.org/a2a”, “version”: “1.0” },
{ “type”: “mcp”, “endpoint”: “https://example.org/mcp” }
],
“status”: “active”
}
Federated Bootstrap Discovery
Organisations expose agents on external domains using an RFC 8615-inspired well-known path structure:
Discovery & Capability Model
Discovery services combine deterministic metadata filtering with semantic matching, ensuring trust and policy rules are strictly enforced over natural language matching.
Controlled Capability Ontologies
planning.permission.guidance
transport.public.journey-planning
environment.flood.risk
education.skills.discovery
tourism.itinerary.planning
Machine Discovery Payload Example
POST /v0.1/discover
{
“intent”: “find-business-grant”,
“jurisdiction”: “Wales”,
“language”: “cy”,
“minimum_trust”: “organisation-verified”
}
// Discovery Response
{
“matches”: [
{
“id”: “wales.1:business/funding”,
“canonical”: “https://id.aiagent.wales/business/funding”,
“match”: 0.97
}
]
}
Trust, Security & Conformance
Trust is expressed as machine-readable attestations rather than subjective quality scores, incorporating a zero-trust posture for inter-agent communication.
The v0.1 Trust Ladder
| Level | Trust Designation | Validation Meaning |
|---|---|---|
| 0 | DISCOVERED | Entry exists in graph, unverified claims. |
| 1 | REGISTERED | Canonical entry registered in system. |
| 2 | IDENTITY VERIFIED | Responsible party identity validated. |
| 3 | ORGANISATION VERIFIED | Confirmed controlled by claimed organisation. |
| 4 | CAPABILITY ATTESTED | Capabilities supported by recognised evidence. |
| 5 | ASSURED / PUBLIC SERVICE | Passed formal public assurance audit. |
Ecosystem Conformance Levels
| Level | Conformance Tier | Requirement Criteria |
|---|---|---|
| Level A | Discoverable Agent | Provides canonical identity and minimum manifest. |
| Level B | Interoperable Agent | Supports standard protocols (A2A, MCP, HTTPS). |
| Level C | Trusted Agent | Carries identity or organisation verification. |
| Level D | Team-Capable Agent | Participates within Wales.Team compositions. |
| Level E | Assured Agent | Meets public service governance & accessibility audits. |
Human Journey & Discovery Execution
A human prompt initiates systemic intent resolution without exposing underlying technical complexity.
Human Intent Capture
User submits natural language prompt to AIAgent.Wales (e.g., “I want to start a food business in Swansea”).
Context & Constraint Resolution
System interprets intent, isolates geography (Swansea), selects preferred language (Welsh/English), and sets trust filters.
Discovery Graph Query
Graph identifies appropriate composite team: wales.team:business/start-food-business.
Team Agent Orchestration
Coordinated execution across funding, planning, premises, food-regulation, and skills agents.
Reference Architecture Topology
End-to-end structural interaction from end users to distributed agent execution environments:
| PEOPLE |
| Businesses / Citizens / |
| Public Bodies / Visitors |
+————+————-+
|
v
+————————–+
| AIAgent.Wales |
| HUMAN DISCOVERY STACK |
+————+————-+
|
v
+——————————–+
| WALES DISCOVERY GRAPH |
| Identity * Capability * Trust |
| Language * Policy * Geography |
+—————+—————-+
|
+————–+————–+
| |
v v
+——————–+ +——————-+
| Wales.Team | | Wales.1 |
| TEAM IDENTITIES | | AGENT IDENTITIES |
+———-+———+ +———+———+
| |
+————–+————+
|
v
+———————–+
| INTERACTION LAYER |
| A2A * MCP * HTTPS |
+———–+———–+
|
v
+——————————+
| DISTRIBUTED AGENT RUNTIMES |
| Public / Private / Cloud |
+——————————+
Core Strategic Outcomes & Vision
Discoverable AI Jurisdiction: Public services discoverable directly by autonomous AI agents.
Bilingual AI Testbed: Full metadata-level Welsh and English operational parity.
Multi-Agent Laboratory: Assembles public services around citizen intent rather than organizational silos.
WALES AGENT INFRASTRUCTURE COUNCIL
AIAgent.Wales • Wales.Team • Wales.1
A discoverable, bilingual, trusted, and interoperable AI-agent ecosystem for Wales.