Giskard Memory
Give your coding agent durable semantic memory across sessions without self-hosting a vector database.
Overview
Giskard Memory is an MCP server for the Build phase that provides pay-per-use semantic memory for AI agents over SSE with SHA256 commitments, Ed25519 signatures, and Lightning billing.
What is this MCP server?
- Remote SSE MCP endpoint at memory.rgiskard.xyz for drop-in agent registration
- Pay-per-use semantic memory with SHA256 commitments for integrity of stored facts
- Ed25519 signatures on operations for verifiable agent-server interactions
- Lightning Network micropayments instead of a flat SaaS subscription for sparse usage
- Version 1.0.1 server published from giskard09/giskard-memory on GitHub
- Server version 1.0.1 per MCP server manifest
- Single remote transport: SSE at https://memory.rgiskard.xyz/sse
- Uses SHA256 commitment, Ed25519 signature, and Lightning payment primitives per description
What problem does it solve?
Agents forget everything when a session ends, and solo builders rarely want to host and bill a full memory stack just to persist a few semantic facts.
Who is it for?
Indie builders running Claude Code or Cursor agents who need cross-session recall and accept Lightning micropayments for occasional memory reads and writes.
Skip if: Teams that require fully on-prem memory, flat monthly pricing only, or a general-purpose application database with SQL guarantees.
What do I get? / Deliverables
After you register the remote MCP server, your agent can read and write durable semantic memory on demand with per-call Lightning payment instead of maintaining your own vector store.
- Registered remote Giskard Memory MCP connection over SSE
- Agent-accessible semantic store with integrity commitments on written content
- Usage-based memory billing via Lightning instead of fixed infra rent
Recommended MCP Servers
Journey fit
Persistent agent memory is wired during product and agent stack setup, before you ship workflows that must remember users, decisions, or prior runs. Agent-tooling is the canonical shelf for MCP servers that extend what Claude Code, Cursor, and similar clients can recall and retrieve at runtime.
How it compares
Hosted semantic-memory MCP integration, not an agent skill that teaches brainstorming or planning rituals.
Common Questions / FAQ
Who is Giskard Memory for?
It is for solo and indie builders who ship AI agents or long agent sessions and want semantic recall without self-hosting embeddings infrastructure.
When should I use Giskard Memory?
Use it during Build when you are wiring agent tooling and need persistent memory across chats, tools, or lightweight production pilots.
How do I add Giskard Memory to my agent?
Add the remote SSE MCP server URL https://memory.rgiskard.xyz/sse in Claude Code, Cursor, or another MCP-capable client, then fund Lightning payments per the server’s billing flow.