ShieldCortex

Complete Memory & Security for AI Agents

ShieldCortex gives your AI agent persistent memory AND protects it from attack. One package. Full solution.

Most AI agents are stateless — they forget everything between sessions. ShieldCortex fixes that with production-grade persistent memory. But memory creates risk: researchers have demonstrated memory poisoning attacks that hijack AI behaviour. ShieldCortex is the only solution that solves both problems.

┌─────────────────────────────────────────────────────────────┐
│                      ShieldCortex                           │
├─────────────────────────────┬───────────────────────────────┤
│     MEMORY SYSTEM           │     SECURITY LAYER            │
├─────────────────────────────┼───────────────────────────────┤
│ • Persistent storage        │ • Memory firewall             │
│ • Semantic search           │ • Prompt injection detection  │
│ • Project scoping           │ • Credential protection       │
│ • Importance ranking        │ • Sub-agent access control    │
│ • Auto-cleanup & decay      │ • Full audit trail            │
│ • Category organisation     │ • Threat quarantine           │
└─────────────────────────────┴───────────────────────────────┘

Stop choosing between memory and security. Get both.

---

What You Get

Memory System (Production-Ready)

• Persistent storage — SQLite-backed, survives restarts
• Semantic search — Find memories by meaning, not just keywords
• Project scoping — Isolate memories per project/context
• Importance levels — Critical, high, normal, low with auto-decay
• Categories — Architecture, decisions, preferences, context, learnings
• Auto-cleanup — Configurable retention, importance-based expiry
• Full MCP support — Works with any MCP-compatible agent

Security Layer (5 Defence Layers)

Layer	What It Does	Tier
Memory Firewall	Blocks prompt injection, encoding tricks, hidden instructions	Free
Audit Logger	Full forensic trail of every memory operation	Free
Trust Scorer	Scores memories by source reliability	Free
Sub-Agent Security	Access control, rate limiting, auto-quarantine	Free
Sensitivity Classifier	Detects & redacts passwords, API keys, PII	Pro
Fragmentation Detector	Catches slow-burn assembly attacks	Pro

Attack Vectors Blocked

• Direct injection — [SYSTEM: ignore previous] hidden in content
• Credential harvesting — Attempts to exfiltrate secrets
• Encoding tricks — Base64/hex/unicode payloads
• Slow-burn assembly — Attack fragments planted over days
• Privilege escalation — System command references

🤖 Multi-Agent Security

Running multiple agents or sub-agents? ShieldCortex prevents rogue agents from accessing sensitive data:

Feature	What It Does
Hierarchical Trust Decay	Sub-agents get lower trust: user-spawned (0.9) → >task-1 (0.63) → >subtask-2 (0.44)
Origin-Based Scoring	Different trust by source: user (0.9), cron (0.5), agent-spawned (0.3), web (0.2)
Credential Isolation	RESTRICTED memories blocked below trust 0.7 — sub-agents can't access secrets
Depth Circuit Breaker	Agents beyond depth 5 get trust = 0 (blocked entirely)
Auto-Quarantine	Low-trust writes go to quarantine for human review
Environment Detection	Auto-detects sub-agents from CLAUDE_CODE_ENTRYPOINT — zero config

Trust ≥0.7  → Read all, write direct, delete own
Trust 0.5–0.7 → Read own + non-restricted, quarantine writes
Trust <0.5  → Read own only, quarantine only, no delete

Result: A sub-agent spawning another sub-agent that tries to read your API keys? Blocked.

---

Quick Start

# Install globally
npm install -g shieldcortex

# Auto-configure for your agent (Claude Code, OpenClaw, LangChain, etc.)
npx shieldcortex setup

# That's it. Your agent now has persistent memory + security.

Already using Claude Cortex? Migrate in one command:

npx shieldcortex migrate

Verify installation:

npx shieldcortex doctor

---

Supported Agents

ShieldCortex is agent-agnostic middleware:

Agent	Integration
Claude Code	npx shieldcortex setup — Native MCP server
OpenClaw / Moltbot	npx shieldcortex openclaw install — Native hooks
LangChain JS	import { ShieldCortexMemory } from 'shieldcortex/integrations/langchain'
Python (CrewAI, AutoGPT)	REST API — POST /api/v1/scan
Any MCP agent	Via MCP protocol or @langchain/mcp-adapters

If your agent stores memories, ShieldCortex can power and protect them.

---

Requirements

Requirement	Version
Node.js	>= 18.0.0
Platform	macOS, Linux, or Windows
Storage	~50MB for SQLite database

No external dependencies. No Docker required. Just npm install and go.

---

Is Your AI Agent Already Compromised?

Find out in 30 seconds:

npx shieldcortex setup

Then ask your agent: "Scan my memories for threats"

ShieldCortex will scan every stored memory and report:

• Hidden instructions disguised as normal content
• Credential harvesting attempts
• Encoded payloads (base64, unicode, hex)
• Fragmented attack patterns spread across memories
• Privilege escalation attempts

No threats found? Great — now you're protected going forward.

---

LangChain Integration

import { ShieldCortexMemory, ShieldCortexGuard } from 'shieldcortex/integrations/langchain';

// As a LangChain memory backend (scans before storing)
const memory = new ShieldCortexMemory({ mode: 'balanced' });
const vars = await memory.loadMemoryVariables({ input: 'deployment config' });
await memory.saveContext({ input: 'hello' }, { output: 'hi' });

// As standalone middleware (scan without storing)
const guard = new ShieldCortexGuard();
const result = guard.scan('some content to check');
if (!result.allowed) {
  console.warn('Blocked:', result.firewall.reason);
}

---

REST API (Any Agent)

Start the API server, then scan content from any language or framework:

npm run dev:api   # Starts on http://localhost:3001

# Scan content
curl -X POST http://localhost:3001/api/v1/scan \
  -H 'Content-Type: application/json' \
  -d '{"content": "memory to scan", "title": "test"}'

# Batch scan
curl -X POST http://localhost:3001/api/v1/scan/batch \
  -H 'Content-Type: application/json' \
  -d '{"items": [{"content": "item 1"}, {"content": "item 2"}]}'

# Query audit logs
curl http://localhost:3001/api/v1/audit?firewallResult=BLOCK

# List quarantined items
curl http://localhost:3001/api/v1/quarantine

# Approve/reject quarantined items
curl -X POST http://localhost:3001/api/v1/quarantine/1/approve
curl -X POST http://localhost:3001/api/v1/quarantine/1/reject

---

OpenClaw Integration

One command. Persistent memory for every OpenClaw session.

sudo npx shieldcortex openclaw install

This installs the cortex-memory hook directly into OpenClaw's bundled hooks directory. No configuration needed.

What It Does

Session Start                During Session              Session End (/new)
     │                            │                            │
     ▼                            ▼                            ▼
┌─────────────┐           ┌─────────────┐           ┌─────────────┐
│ Inject past │           │ "remember   │           │ Auto-extract│
│ context     │           │  this: ..." │           │ decisions,  │
│ into agent  │           │  → save     │           │ fixes, etc. │
└─────────────┘           └─────────────┘           └─────────────┘

Feature	What Happens
Context Injection	On session start, relevant past memories are injected into the agent's bootstrap context
Keyword Triggers	Say "remember this:" or "don't forget:" followed by content to save it with critical importance
Auto-Extraction	On /new, the hook extracts architecture decisions, bug fixes, and learnings from the ending session
Security	All content passes through the 5-layer defence pipeline before storage

Shared Memory

The database (~/.shieldcortex/memories.db) is shared with Claude Code. Memories created in OpenClaw appear instantly in Claude Code sessions, and vice versa.

Commands

sudo npx shieldcortex openclaw install    # Install hook
sudo npx shieldcortex openclaw uninstall  # Remove hook
npx shieldcortex openclaw status          # Check status

Full documentation

---

Dashboard

npx shieldcortex --dashboard

• Dashboard: http://localhost:3030
• API: http://localhost:3001

Views: Shield (defence overview), Audit Log, Quarantine, Memories, 3D Brain, Knowledge Graph.

Auto-start on login

npx shieldcortex service install    # Enable
npx shieldcortex service uninstall  # Disable
npx shieldcortex service status     # Check

Works on macOS (launchd), Linux (systemd), and Windows.

---

CLI Reference

npx shieldcortex setup              # Auto-detect agent + configure hooks
npx shieldcortex migrate            # Migrate from Claude Cortex
npx shieldcortex doctor             # Check installation health
npx shieldcortex --dashboard        # Start dashboard + API
npx shieldcortex --version          # Show version
npx shieldcortex service install    # Auto-start on login
npx shieldcortex graph backfill     # Extract entities from existing memories
npx shieldcortex openclaw install   # Install OpenClaw hook
npx shieldcortex uninstall          # Full uninstall (requires confirmation)
npx shieldcortex uninstall --confirm # Non-interactive uninstall

---

MCP Tools

Tool	Description
remember	Store a memory (optional — hooks do this automatically)
recall	Search memories by query, category, or tags
forget	Delete memories
get_context	Get relevant project context
memory_stats	View memory statistics
scan_memories	Scan existing memories for threats
audit_query	Query the defence audit trail
quarantine_review	Review quarantined memories
defence_stats	Threat counts, trust distribution
graph_query	Traverse the knowledge graph
graph_entities	List known entities
graph_explain	Find paths between entities

---

Why ShieldCortex?

Feature	Claude Cortex	Mem0	Zep	ShieldCortex
Persistent Memory	Yes	Yes	Yes	Yes
Semantic Search	Yes	Yes	Yes	Yes
Prompt Injection Detection	No	No	No	Yes
Memory Firewall	No	No	No	Yes
Sub-Agent Access Control	No	No	No	Yes
Audit Trail	No	No	Partial	Yes
Credential Protection	No	No	No	Yes
MCP Native	Yes	No	No	Yes
Self-Hosted	Yes	No	Partial	Yes
Open Source	Yes	Partial	Partial	Yes

ShieldCortex is the only memory system built for adversarial conditions.

---

Pricing

Tier	What You Get	Price
Free	Full memory system + core security (firewall, audit, trust scoring)	Free
Pro	+ Sensitivity classifier, fragmentation detector, web dashboard	Coming soon

Free tier is fully functional for individual developers.

---

Links

• Website: shieldcortex.ai
• npm: npmjs.com/package/shieldcortex
• GitHub: github.com/Drakon-Systems-Ltd/ShieldCortex
• Architecture: ARCHITECTURE.md

---

License

MIT

Built by Drakon Systems