🧠 Synapse Layer CLI

Persistent zero-knowledge memory for AI agents. Store, recall and transfer encrypted context across models.

"Giving Agents a Past. Giving Models a Soul."

Synapse Layer is the universal memory layer for AI agents. It provides persistent, private, model-agnostic memory with zero-knowledge encryption — the server never sees your plaintext data.

🚀 Quick Start

# Install globally
npm install -g synapse-layer

# Or use directly with npx
npx synapse-layer remember "User prefers dark mode" --user user_123

Setup

export SYNAPSE_API_KEY="your-supabase-anon-key"

📖 Commands

synapse remember — Store a memory

Store a memory with zero-knowledge encryption. Content is encrypted locally before being sent.

# Basic usage
synapse remember "User prefers concise responses" --user user_123

# With importance score
synapse remember "API key rotated on March 2026" --user user_123 --importance 0.9

# With category
synapse remember "Prefers TypeScript over Python" --user user_123 --category preference

Options:

Categories: preference, fact, instruction, emotion, goal, relationship, skill, context

synapse recall — Semantic memory search

Search your encrypted memory vault using vector similarity.

# Basic recall
synapse recall "communication preferences" --user user_123

# With limit
synapse recall "what does the user prefer?" --user user_123 --limit 10

# With threshold
synapse recall "project deadlines" --user user_123 --threshold 0.8

Options:

synapse handover — Neural Handover™

Transfer context between AI models seamlessly. Create signed, encrypted handover packages.

# Transfer to GPT
synapse handover --to gpt --user user_123

# Transfer to Claude with custom expiry
synapse handover --to claude --user user_123 --expires 7200

# Transfer to Gemini with more memories
synapse handover --to gemini --user user_123 --limit 20

Supported models: gpt, claude, gemini, llama, mistral, deepseek

Options:

synapse import — Import memory blob

Load a previously exported encrypted memory package into the vault.

# Import from base64 string
synapse import --blob <base64_string> --user user_123

# Import from file
synapse import --blob $(cat handover.b64) --user user_123 --category context

Options:

synapse status — Vault health check

Display current vault state, including health score and memory statistics.

synapse status --user user_123

Options:

synapse forget — Delete memories

Remove memories matching a query. Supports LGPD/GDPR compliance with soft-delete.

# Soft-delete (default, LGPD/GDPR compliant)
synapse forget "old preferences" --user user_123

# Permanent deletion
synapse forget "outdated API keys" --user user_123 --hard

Options:

🔌 MCP Server

Connect any MCP-compatible client to the Synapse Layer server:

https://rbeycxzizrrdmxpilepc.supabase.co/functions/v1/mcp-server

🔒 Zero-Knowledge Architecture

All data is encrypted client-side with AES-256-GCM before leaving your machine. The server stores only encrypted blobs and vector embeddings — it never has access to your plaintext data.

• Encryption: AES-256-GCM with PBKDF2 key derivation (210,000 iterations)
• Embeddings: gte-small (384 dimensions)
• Compliance: LGPD/GDPR ready with soft-delete support
• Signing: HMAC-SHA256 for Neural Handover™ packages

🏗️ Agent Integration Examples

LangChain

from synapse_memory import SynapseMemory

memory = SynapseMemory(api_key="your-key", user_id="user_123")
memory.store("User prefers brief answers", importance=0.8)
results = memory.recall("communication style")

Direct API (fetch)

const response = await fetch('https://rbeycxzizrrdmxpilepc.supabase.co/functions/v1/mcp-server', {
  method: 'POST',
  headers: { 'Content-Type': 'application/json', 'Authorization': 'Bearer YOUR_KEY' },
  body: JSON.stringify({
    jsonrpc: '2.0', id: 1,
    method: 'tools/call',
    params: { name: 'store_memory', arguments: { user_id: 'user_123', /* ... */ } }
  })
});

🌐 Environment Variables

📄 License

Apache 2.0 — Built in São Paulo · synapselayer.org

Keywords

agent persistent memory, cross-model context transfer, zero-knowledge memory, recall past conversation, memory handover, MCP memory tool, neural handover, conflict resolution memory, semantic memory search, agent state persistence, mcp, langchain, ai-agents, persistent memory for AI, encrypted agent memory, model context protocol