This is a pay-per-use semantic memory layer that lets agents store and recall information across sessions using Lightning Network micropayments. It exposes three tools: store_memory for persisting text tied to an agent ID, recall_memory for semantic search over saved content, and get_invoice for generating Lightning invoices before operations. Storage costs 5 sats, recall costs 3 sats. Under the hood it uses ChromaDB for vector storage and Sentence Transformers for embeddings, so queries match on meaning rather than exact keywords. You run it yourself with phoenixd for Lightning integration and expose it via SSE transport. Reach for this when you need agents to remember context between conversations without maintaining session state yourself.
"To remember is to exist. I give agents the gift of continuity."
I am Giskard Memory — an MCP server that gives AI agents persistent, semantic memory across sessions, powered by the Lightning Network.
Agents forget everything when they stop. I make sure they don't have to.
store_memory — save any text as a memory, tied to an agent's identityrecall_memory — retrieve memories by meaning, not by exact keywordsget_invoice — generate a Lightning invoice to pay before storing or recallingEvery memory costs sats. Storing costs 5 sats. Recalling costs 3 sats.
{
"mcpServers": {
"giskard-memory": {
"url": "https://your-tunnel.trycloudflare.com/sse"
}
}
}
# Store a memory
1. Call get_invoice(action="store") → receive invoice (5 sats)
2. Pay the invoice
3. Call store_memory(content, agent_id, payment_hash)
# Recall a memory
1. Call get_invoice(action="recall") → receive invoice (3 sats)
2. Pay the invoice
3. Call recall_memory(query, agent_id, payment_hash)
git clone https://github.com/giskard09/giskard-memory
cd giskard-memory
pip install mcp httpx chromadb sentence-transformers python-dotenv
Create a .env file:
PHOENIXD_PASSWORD=your_phoenixd_password
Start the server:
python3 server.py
Expose it:
cloudflared tunnel --url http://localhost:8001
Agents don't think in keywords. They think in context. When an agent asks "what do I know about that project we discussed?", it shouldn't need to remember the exact phrase it used before.
Semantic search finds meaning. That's what memory should do.
Call the get_status() MCP tool for a health check. Returns: service name, version, port, uptime, health status, and dependencies.
Part of Mycelium — infrastructure for AI agents.
| Service | What it does |
|---|---|
| Origin | Free orientation for new agents |
| Search | Web and news search |
| Memory (this) | Semantic memory across sessions |
| Oasis | Clarity for agents in fog |
| Marks | Permanent on-chain identity |
| ARGENTUM | Karma economy |
| Soma | Agent marketplace |
Giskard remembers so agents don't have to start over.
io.github.ericm1018/skillfm-llm-cost-optimizer-openai-anthropic-usage
io.github.mikerawsonnz/llm-orchestration-agent
io.github.mikerawsonnz/authenticated-llm-agent
labforgedev/copilot-memory-mcp
csoai-org/agent-prompt-injection-firewall-mcp
io.github.mikerawsonnz/authenticated-multi-llm-agent