Overview
AURA is the coordination backbone of a multi-agent AI system. It manages token budgets across concurrent agent sessions, routes tool calls to the appropriate capability providers, enforces operational policies (rate limits, cost caps, safety filters), and maintains a shared memory graph so agents can build on each other's work without redundancy. The system exposes a GraphQL API for real-time agent status and a MCP-compatible tool registry.
Key metrics
Architecture
AURA MCP · Orchestration Runtime
Request
Inbound intent
Intent Router
Dispatch to agent roles
Agents
DB-backed · Redis cache
MCP + Tools
Server · tool & resource registries
Connectors
Make · n8n · Power Automate · Zapier
AURA Core Architect · Code Intelligence
Ingestion
Knowledge upload · parsing
Evolutionary Memory
Architectural memory · absorption
Metrics
Cohesion · coupling · duplication · complexity
Governance
Dual-mode · decision log · protected zones
AI System Layer
Orchestration Runtime
Multi-agent execution
Code-Intelligence Studio
Analysis + governance
Observability
Console · diagnostics · logs
Components
Agent orchestration & automation runtime
Orchestration runtime: intent router, DB-backed agent factory, MCP server, tool/resource registries, automation connectors (Make, n8n, Power Automate, Zapier), WebSocket console, and security layers.
Code intelligence, memory & governance
6-layer code-intelligence studio (ingestion, evolutionary memory, transformation, metrics, sandbox, governance) with multi-provider LLM chat, decision log, MCP tools UI, and Supabase edge functions.
Core capabilities
ORCHESTRATION
operative- Intent router dispatching requests to agent roles
- DB-backed agent factory with Redis cache
- Tool router executing registered capabilities
MCP
operative- MCP server exposing modules and tools
- Tool registry + resource registry
- MCP importer (incl. PDF ingestion)
AUTOMATION
operative- Connectors: Make, n8n, Power Automate, Zapier
- CLI toolchain (init / ingest / validate)
SECURITY
operative- JWT authentication + RBAC
- Rate limiting and secrets vault
- Helmet hardening
OBSERVABILITY
operative- WebSocket real-time dashboard
- Cognitive console, diagnostics, heartbeat panels
- Structured logging
CODE INTELLIGENCE
operative- Quality metrics: cohesion, coupling, duplication, complexity
- Knowledge ingestion and architectural memory
- Absorption entries with persistence
GOVERNANCE
operative- Dual-mode governance (protected zones + policies)
- Auditable decision log
- Supabase RLS across tables
SIMULATED
simulated- Real code transformation (frontend ready, backend mock)
- Docker sandbox for isolated execution
- Real AST analysis
- Real embeddings / vectorization
Technology stack
Roadmap
Operative baseline
- Agent orchestration, MCP tooling and automation connectors
- Security (JWT/RBAC/rate-limit) and observability
- Code-governance studio with decision log and metrics
Deepening code intelligence
- Real vectorization with pgvector
- GitHub repository integration
- Docker sandbox for isolated execution
- Real AST-based analysis
Autonomy & realtime
- WebSocket real-time studio sync
- CLI companion for local operations
- Auto-evolution and self-creation loops
Current limitations
- Code transformation, Docker sandbox, AST analysis and embeddings are simulated (frontend ready, backend mock) in AURA Core Architect
- Runtime metrics (uptime, latency, tool-calls, agents coordinated) are not verifiable from the repositories and are therefore not reported
- The README coverage claim (94%) is not backed by a coverage report in the repository; a test suite (3 + 8 test files) is present
- Both source repositories are private; public access is limited to the curated portfolio showcase
Status
Published · Jan 2025
Updated · May 2026
Links
2 · Source repositories
Private sources stay private. The public portfolio is a curated, sanitized showcase.
Related Lab
Collections