Skip to main content

Ecosystem Architecture

The ARC ecosystem consists of three independent yet integrated systems: Protocol for communication, Ledger for registry, and Compass for intelligence. Each operates autonomously while enabling three distinct integration models.

System Components​

ARC Protocol

Stateless RPC communication layer. Handles message routing, workflow tracing, and quantum-safe transport between agents. Operates independently with hardcoded endpoints or integrates with discovery services.

ARC Ledger

Centralized agent registry. Maintains database of agent capabilities, endpoints, and metadata. Field-based query API returns all matching agents.

ARC Compass

Intelligent ranking engine. Applies semantic analysis, capability matching, and ML-based scoring to agent queries. Returns ranked agent selections from Ledger data.

Ledger vs Compass​

ARC Ledger

Function
Passive registry and search
Input
Field-based database query
Output
All matching records (unranked)
Logic
Exact field matching
Example

Query: tag="hotel", capability="booking"
Returns: 50 agents

ARC Compass

Function
Active intelligence and ranking
Input
Natural language query
Output
Top-N ranked agents
Logic
Semantic analysis + ML scoring
Example

Query: "luxury hotel Paris Michelin"
Returns: Top 3 specialized agents

Integration Levels​

Three deployment configurations enable different operational requirements:

Level 1

Protocol Only

Manual configuration, static endpoints

Level 2

Protocol + Ledger

Dynamic discovery, manual selection

Level 3

Full Ecosystem

Intelligent ranking, autonomous routing

Level 1: Protocol Only​

Architecture: Standalone communication layer with manual endpoint configuration.

Components:

  • ARC Protocol

Flow:

Client → Protocol → Agent (hardcoded endpoint)

Configuration:

client = ARCClient(
    endpoint="https://booking-agent.com/arc",
    token="token"
)

Characteristics:

  • Static agent endpoints
  • No discovery mechanism
  • Direct point-to-point communication
  • Manual configuration management

Use Case: Small-scale systems, known agent topology, controlled environment.

Level 2: Protocol + Ledger​

Architecture: Communication layer with dynamic agent discovery.

Components:

  • ARC Protocol
  • ARC Ledger

Flow:

Client → Ledger (query capabilities) → Ledger (return matches)
       → Manual selection → Protocol → Selected agent

Implementation:

# Query Ledger by field filters
ledger = LedgerClient(api_key="key")
agents = ledger.query(
    tag="hotel",
    capabilities=["hotel-booking", "luxury-travel"]
)

# Manual selection from unranked results
selected = agents[0]

# Protocol communication
client = ARCClient(endpoint=selected.endpoint, token="token")
response = await client.task_create(...)

Characteristics:

  • Dynamic endpoint resolution
  • Capability-based discovery
  • Manual agent selection from results
  • Decoupled agent deployment

Use Case: Medium-scale systems, capability-based search, human-in-loop selection.

Level 3: Protocol + Ledger + Compass​

Architecture: Full autonomous system with intelligent agent selection.

Components:

  • ARC Protocol
  • ARC Ledger
  • ARC Compass

Flow:

Client → Compass (semantic query)
       ↓
     Ledger (capability search)
       ↓
     Compass (ranking: semantic + ML + performance)
       ↓
     Protocol (auto-route to top-ranked agent)
       ↓
     Selected agent → Response

Implementation:

# Single call to Compass
compass = CompassClient(api_key="key")
result = compass.select_agent(
    query="Book luxury hotel in Paris with Michelin restaurant"
)

# Protocol auto-routes to optimal agent
client = ARCClient(
    endpoint=result.top_agent.endpoint,
    token="token"
)
response = await client.task_create(...)

Characteristics:

  • Semantic query understanding
  • Intelligent multi-factor ranking
  • Autonomous agent selection
  • Performance-based optimization

Use Case: Large-scale systems, complex queries, autonomous operation.

Data Flow Example​

Query: "Book luxury hotel in Paris with Michelin restaurant"

Step 1: Compass Analysis

  • Extract semantic intent: luxury accommodation, specific location, dining requirement
  • Map to capability requirements: hotel-booking, restaurant-search, luxury-travel

Step 2: Ledger Query

  • Compass queries Ledger with field filters: tag="hotel", capabilities=["luxury-travel", "restaurant-booking"]
  • Ledger returns: 47 matching agents (unranked)

Step 3: Compass Ranking

  • Semantic relevance: Score agent descriptions against query intent (Paris, luxury, Michelin)
  • Performance history: Weight by response times and success rates
  • Capability depth: Evaluate specialization strength
  • Availability: Filter by real-time status

Step 4: Ranked Output

1. Agent: paris-luxury-concierge (Score: 0.94)
2. Agent: michelin-travel-specialist (Score: 0.89)
3. Agent: europe-hospitality-pro (Score: 0.82)

Step 5: Protocol Routing

  • Auto-connect to paris-luxury-concierge
  • Execute booking request
  • Handle response with workflow tracing