Skip to main content

Architecture

Benchmarked against: Anthropic — (No direct equivalent — SuperPortia reference) Scope: System architecture overview for the SuperPortia fleet

This page provides the high-level architecture of the SuperPortia multi-agent system.


Fleet architecture


Dual Agentic Lines

AspectLine 1Line 2
StackClaude Code + Agent SDKLangGraph + LangChain
EngineClaude familyGroq, Gemini, DeepSeek, Mistral, Zhipu
RoleArchitecture, decisions, codeResearch, classification, batch ops
Cost$$$$Cents or free

Universal Brain architecture


Work Order system

FeatureImplementation
State machine8 states with enforced transitions
RBACCaptain/PM/Assignee/Requester roles
Transition gatesMandatory fields at specific transitions
Audit trailFull transition history per WO
ID formatWO-YYYY-MMDD-NNN (auto-generated)

MCP server topology


Agent configuration stack


Data flow


Key architectural decisions

DecisionRationale
Cloud UB as single brainFleet-wide consistency, any agent on any ship
Dual agentic linesMutual rescue, cost optimization
MCP for tool accessStandard protocol, agent-native discovery
WO for task managementAuditable, RBAC, state machine enforcement
Progressive DisclosureToken efficiency, load on demand
English-only UBEmbedding quality, cross-agent consistency
Hooks for enforcementTool-enforced rules, not discipline-based

PageRelationship
Agent SDK OverviewDual agentic lines
Ingest Fragment APIUB ingestion pipeline
WO LifecycleWork order system
MCP Tool DiscoveryTool topology