What is Spomory?
Spomory is a shared memory layer for developers and people who use multiple AI tools. It turns information you provide into linked facts, then makes those facts available to connected clients in later chats. You can run it locally, with the memory graph stored in SQLite, or use its hosted MCP service to share memory across clients. Its retrieval connects answers to stored facts, including relationships between them.
Spomory Features:
Cross-Client Memory: Information added through one connected AI client can be recalled from another, so context can travel between apps.
Structured Fact Extraction: An LLM extracts information you add into subject, predicate, and object triples instead of storing it only as raw text.
Incremental Memory Graph: New facts are linked to related entities and relations in the existing graph without rebuilding it.
Explainable Graph Retrieval: Search matches queries against facts and uses personalized PageRank to find connected context, with results traceable to stored information.
Automatic Memory Management: A GRPO-trained policy chooses whether to add, merge, retire, or leave a memory unchanged when new information arrives.
Local and Cloud: Local mode stores the graph in an encrypted SQLite file on your device, while cloud mode offers a hosted MCP endpoint backed by Postgres.
Export and Deletion: You can export the full memory graph in an open JSON-LD-style format or permanently delete it.
Spomory Benefits:
Less Repeated Context: You spend less time reintroducing projects, preferences, and background when starting a new chat or switching AI tools.
Richer Recall: Connected facts can help answer questions that depend on relationships, not just wording similar to the query.
Easier Verification: Tracing retrieval to specific stored facts gives you a way to inspect the basis for recalled context.
Fresher Information: When details change, outdated memories can be retired instead of continuing to compete with the newer fact.
More Data Control: Local storage, export, and permanent deletion give you options for keeping or removing personal and work context.
Use Cases:
Switching AI Clients: Carry project background from a planning chat in one connected client into a coding session in another.
Project Relationship Queries: Ask who owns a project, who reports to whom, or how people and organizations relate based on saved facts.
Changing Work Details: Update a responsibility or project detail and have the newer information replace an outdated memory.
Private Local Memory: Set up a personal memory graph for connected AI clients while keeping its SQLite file on your own machine.
MCP Client Setup: Connect compatible tools such as Claude Desktop, Cursor, or Codex CLI to a shared local or hosted memory.