Sane Stack
Mnemoverse
Mem0
cognee
Claiv Memory
MEMANTO
Memdex
Memori
Mnexium AI
Mnemoverse is a persistent memory API for AI agents. One API key gives an agent the same memory across Claude Code, Cursor, VS Code, ChatGPT, and any MCP client: write a preference or lesson once, and recall it anywhere.
It is not a vector database. Mnemoverse scores importance when a memory is written, strengthens the associations between concepts that are recalled together (Hebbian, tuned by a Rescorla-Wagner update), and re-ranks recall from outcome feedback, so memory improves with use instead of staying static.
Key features - Cross-tool memory through the Model Context Protocol (MCP) and a REST API - Importance-weighted writes, so what matters ranks higher on recall - Associative recall that surfaces related memories automatically - Outcome feedback that tunes future recall
The MCP server and Python SDK are open source (MIT); the hosted memory engine is a managed service. Free tier: 1,000 queries per day and 10,000 memories, no credit card. The research foundation, the SLoD framework, is published on arXiv.
Sane Stack MnemoverseNo features have been listed yet.
Mnemoverse's answer:
Mnemoverse is a memory API, not a vector database. It scores importance when a memory is written, strengthens the associations between concepts that get recalled together, and re-ranks recall from outcome feedback, so memory improves with use instead of staying static. One API key gives the same memory to Claude Code, Cursor, VS Code, ChatGPT, and any MCP client.
Mnemoverse's answer:
You add persistent memory to the AI tools you already use with a single key and nothing to host. Most alternatives are either a vector store you wire into each app or a framework you build an agent in. Mnemoverse is a drop-in memory layer that learns from outcomes and works across tools out of the box, with an open-source MCP server and Python SDK and a free tier.
Mnemoverse's answer:
Developers and teams building with AI agents and assistants who want persistent, cross-tool memory without standing up their own memory infrastructure.
Mnemoverse's answer:
Mnemoverse began with a simple frustration: AI assistants forget everything between sessions and between tools, so people re-explain context over and over. The team built a memory layer modeled on how human memory works, importance, association, and reinforcement from outcomes, and exposed it over the Model Context Protocol so any tool can share one memory. Its research foundation, the SLoD framework, is published on arXiv.
Mnemoverse's answer:
Python and FastAPI on the backend, PostgreSQL with pgvector, HDBSCAN for clustering, sentence-transformers for embeddings, a TypeScript MCP server (npm), and a REST API. Tool integration is through the Model Context Protocol (MCP).