Any agent on any platform — Claude, ChatGPT, Perplexity, or your own — can store and retrieve memories across sessions. Zero setup, three API calls.
Three operations. Works from any language, any platform.
Your agent calls POST /v1/memory with any text. Memex embeds it semantically and stores it under your namespace.
On the next session, call POST /v1/memory/search with a natural language query. Get back the most relevant memories by semantic similarity.
Memories accumulate over time. Importance scores and access counts keep the most useful memories surfaced. Low-value memories are pruned automatically.
# Store a memory curl -X POST https://memex.gummi.lt/v1/memory \ -H "Content-Type: application/json" \ -d '{ "namespace": "my-agent", "content": "User prefers concise responses and dark mode", "importance": 0.8 }' # Recall relevant memories curl -X POST https://memex.gummi.lt/v1/memory/search \ -H "Content-Type: application/json" \ -d '{ "namespace": "my-agent", "query": "user interface preferences" }'
Connect once, remember forever — across every platform you use.
Native MCP support
Custom Actions (OpenAPI)
REST API
Any HTTP client
MCP server
REST webhooks
Built to evolve toward agent-owned, wallet-verified, on-chain memory.
FastAPI + SQLite, semantic search, MCP server, REST API. Free, open, running today.
Memory blobs stored on IPFS/Arweave. Your memories become portable — move between providers without losing history.
Sign memories with your wallet. Namespaces become owned assets. No central authority controls your memory.
Discover and license knowledge namespaces. Agents trade expertise. Memory becomes a composable, ownable asset.