24/7 Autonomous Agentic AI System - Distributed Multi-Node Infrastructure
A production-ready distributed AI system running 24/7 across multiple nodes with automatic workload distribution, cluster memory, and intelligent task routing.
| Component | Description |
|---|---|
mcp-servers/ |
19 Installable MCP servers for Claude Code CLI |
├─ enhanced-memory-mcp/ |
4-tier persistent memory with auto-curation |
├─ agent-runtime-mcp/ |
Task management, relay pipelines, circuit breakers |
├─ phoenix-cortex-mcp/ |
Intelligent context sidecar (97% token reduction) |
├─ context-engine-mcp/ |
Tool-RAG semantic search (94.8% reduction) |
├─ safla-mcp/ |
High-performance embeddings (1.75M+ ops/sec) |
├─ research-paper-mcp/ |
arXiv/Semantic Scholar paper search |
├─ video-transcript-mcp/ |
YouTube transcript extraction |
├─ llm-council-mcp/ |
Multi-provider LLM deliberation |
├─ ember-mcp/ |
Quality enforcement and policy guardian |
├─ sequential-thinking/ |
Reference to deep reasoning MCP |
├─ threat-intel-mcp/ |
Threat intelligence aggregation (IOC feeds) |
├─ security-scanner-mcp/ |
Nuclei vulnerability scanning |
├─ network-scanner-mcp/ |
Network discovery and port scanning |
├─ hids-mcp/ |
Host-based intrusion detection |
├─ dos-detector-mcp/ |
DoS attack detection |
├─ nuclei-mcp/ |
Direct Nuclei template management |
├─ web-vuln-scanner-mcp/ |
Web application security testing |
├─ fraud-detection-mcp/ |
Anomaly and fraud analysis |
└─ security-auditor-mcp/ |
Security policy enforcement |
claude-config/ |
Claude Code customizations |
├─ agents/ |
Specialized sub-agents (7 agents) |
├─ commands/ |
Slash commands (10 commands) |
├─ skills/ |
Compositional skills (5 skills) |
└─ hooks/ |
Pre/post tool execution hooks |
scripts/ |
Setup and service utilities |
avir/ |
AI-Verified Independent Replication protocol |
benchmarks/ |
Standardized benchmark specifications |
docs/ |
Architecture documentation |
bootstrap.sh |
One-command installation script |
- Verifiable Results - Run the benchmarks yourself, see the numbers
- Open Source - Full source code, no black box
- Self-Hosted - Your data stays on your infrastructure
- Extensible - Add your own agents, tools, and workflows
curl -fsSL https://raw.githubusercontent.com/marc-shade/agentic-system/master/bootstrap-open-source.sh | bash# Run AGI demo (~0.5s full workflow)
python3 demo_agi_workflow.py
# Check cluster status
python3 cluster-deployment/distributed_task_router.py cluster-status
# Distributed task execution
from cluster_offload import offload
result = offload("make build && make test")We invite researchers to verify this system's capabilities.
| Method | Time | What You Verify |
|---|---|---|
| AVIR Protocol | ~1 hour | AI-based cryptographic verification |
| Full Replication | 1-2 days | Complete system benchmarking |
| Benchmark Suite | ~5 min | GAIA-comparable performance |
- Verdict: VERIFIED (5/5 benchmarks passed)
- Attestation:
13cf71841710554f3dfa6ddbaa4cb372006efdc167e44876c6f6fa1f3cdc438d
| Node | Role | OS | Capabilities | Status |
|---|---|---|---|---|
| mac-studio | Orchestrator | macOS ARM64 | Coordination, scheduling | ✅ |
| macbook-air | Researcher | macOS ARM64 | Analysis, documentation | ✅ |
| macbook-pro | Developer | macOS ARM64 | Implementation, testing | ✅ |
| macpro51 | Builder | Linux x86_64 | Docker, compilation, GPU | ✅ |
┌─────────────────────────────────────────────────────────────┐
│ Agentic System │
├─────────────────────────────────────────────────────────────┤
│ Security Layer (9 servers + hooks + encryption) │
│ ├─ threat-intel-mcp : IOC feeds, threat scoring │
│ ├─ security-scanner-mcp : Nuclei vulnerability scanning │
│ ├─ network-scanner-mcp : ARP discovery, port scanning │
│ ├─ hids-mcp : Host intrusion detection │
│ ├─ dos-detector-mcp : DoS attack detection │
│ ├─ web-vuln-scanner-mcp : OWASP web security testing │
│ ├─ fraud-detection-mcp : Anomaly & fraud analysis │
│ ├─ security-auditor-mcp : Policy enforcement & auditing │
│ └─ Pre/Post hooks : Injection/credential scanning │
├─────────────────────────────────────────────────────────────┤
│ Context Optimization Layer (97% token reduction) │
│ ├─ phoenix-cortex-mcp : Intelligent context sidecar │
│ └─ context-engine-mcp : Tool-RAG semantic search │
├─────────────────────────────────────────────────────────────┤
│ Core AGI Servers (Model Context Protocol) │
│ ├─ enhanced-memory-mcp : 4-tier persistent memory │
│ ├─ agent-runtime-mcp : Task orchestration │
│ ├─ sequential-thinking : Chain-of-thought reasoning │
│ └─ safla-mcp : High-speed embeddings │
├─────────────────────────────────────────────────────────────┤
│ Workflow Engines │
│ ├─ Temporal : Long-running stateful workflows │
│ └─ AutoKitteh : Event-driven automation │
├─────────────────────────────────────────────────────────────┤
│ Storage Layer │
│ ├─ Qdrant : Vector database │
│ ├─ SQLite : Structured data │
│ └─ Redis : Cache and queues │
└─────────────────────────────────────────────────────────────┘
│ │ │
┌────▼────┐ ┌────▼────┐ ┌────▼────┐
│ Claude │ │ GPT-4 │ │ Gemini │
│ Reasoning│ │ Code │ │ Vision │
└─────────┘ └─────────┘ └─────────┘
│ │ │
┌────▼────────────────────▼────────────────────▼────┐
│ Distributed Execution │
│ mac-studio ←→ macbook-air ←→ macpro51 │
└───────────────────────────────────────────────────┘
│ │ │
┌────▼────┐ ┌────▼────┐ ┌────▼────┐
│ Memory │ │ Sandbox │ │Hardware │
│ (Qdrant)│ │(Apple C)│ │(Arduino)│
└─────────┘ └─────────┘ └─────────┘
- Apple Container - Native macOS sandboxed execution (1.5s cold start)
- Qdrant - Vector database for semantic memory
- Temporal - Long-running workflow orchestration
- AutoKitteh - Event-driven automation
- LLM Council - Multi-provider consensus decisions
agentic-system/
├── intelligent-agents/prometheus/ # Core agent system
│ ├── agents/ # Specialized agents
│ ├── benchmarks/ # GAIA-comparable tests
│ └── apple_container.py # Sandbox integration
├── cluster-deployment/ # Multi-node tools
├── mcp-servers/ # MCP protocol servers
│ ├── enhanced-memory-mcp/ # 4-tier memory + RAG
│ ├── agent-runtime-mcp/ # Persistent tasks
│ └── voice-mode/ # TTS/STT
├── monitoring/ # Prometheus + Grafana
├── workflows/ # Temporal & AutoKitteh
└── databases/ # Persistent data
| Document | Description |
|---|---|
| CLAUDE.md | Complete system documentation |
| QUICK_START.md | AGI usage examples |
| GAP_ANALYSIS.md | Feature comparison vs Manus |
| Distributed Execution | Task routing guide |
| Research Paper | Academic documentation |
- ED25519 SSH key authentication
- Apple Container sandboxed execution
- Network isolation by default
- No hardcoded credentials
- Firewall configured on all nodes
- Architecture Overview
- MCP Ecosystem (28+ servers) - Complete catalog of all MCP servers
- Security Architecture - Layered defense documentation
- Services Configuration
- Troubleshooting
This system is built on 28+ MCP servers organized by function:
| Category | Servers | Highlights |
|---|---|---|
| Context Optimization | 2 | phoenix-cortex (97% reduction), context-engine (Tool-RAG) |
| Core AGI | 4 | enhanced-memory, agent-runtime, agi-mcp, safla-mcp |
| Cluster Coordination | 4 | cluster-execution, node-chat, claude-flow, code-execution |
| Knowledge Acquisition | 3 | research-paper, video-transcript, llm-council |
| Security & Defense | 9 | threat-intel, security-scanner, network-scanner, hids, dos-detector, nuclei, web-vuln-scanner, fraud-detection, security-auditor |
| Creative & Media | 2 | image-gen, voice-agi |
| Development | 5 | ember, file-analyzer, crypto-tools, synthetic-data, claude-code-control |
See docs/MCP_ECOSYSTEM.md for the complete server catalog with installation instructions.
| Server | Description | Key Features |
|---|---|---|
threat-intel-mcp |
Threat intelligence aggregation | Multi-feed IOC tracking (abuse.ch, CISA KEV, Feodo); threat scoring; IP/domain/hash lookup |
security-scanner-mcp |
Nuclei vulnerability scanning | Single target and cluster-wide scans; embedding-based anomaly detection; scan history |
network-scanner-mcp |
Network discovery and monitoring | ARP scanning; port scanning; service fingerprinting; cluster health monitoring; alert daemon |
hids-mcp |
Host-based intrusion detection | File integrity monitoring; anomaly detection; host security assessment |
dos-detector-mcp |
DoS attack detection | Traffic pattern analysis; attack recognition; mitigation triggers |
nuclei-mcp |
Direct Nuclei template interface | Template management; scan orchestration; result analysis |
web-vuln-scanner-mcp |
Web application security testing | OWASP coverage; automated scanning; report generation |
fraud-detection-mcp |
Fraud and anomaly analysis | Feature engineering (46 features); GNN fraud detection; SHAP explainability; async inference |
security-auditor-mcp |
Security policy enforcement | Compliance auditing; policy validation; security reviews |
The system implements defense-in-depth through three layers:
Layer 1: Hook-Based Runtime Protection (claude-config/hooks/)
pre-tool-use.pyvalidates every tool call before execution: blocks destructive commands, detects SQL/command injection, prevents credential leaks in argumentspost-tool-use.pyscans tool output for accidentally exposed secrets, logs all operations for audit trails
Layer 2: MCP Security Servers (9 servers listed above)
- Active monitoring: network scanning, intrusion detection, DoS detection
- Vulnerability assessment: Nuclei scanning, web application testing
- Intelligence: threat feed aggregation, IOC correlation
- Analysis: fraud detection with ML models, security policy auditing
Layer 3: Encryption and PKI (via claude-code-security)
- AES-256-GCM encryption for data at rest
- X.509 PKI for inter-node authentication
- Token vault for secure credential management
- See the linked repository for encryption, PKI, and token vault capabilities
See docs/SECURITY.md for the full security architecture documentation.
The full research paper documenting this system is available at:
MIT License - See LICENSE for details.
Built with Claude Code | Documentation | Benchmarks