Mnemoverse

Persistent, cross-tool memory for AI agents.

Mnemoverse gives agents a shared memory layer they can use across Cursor, Claude Code, VS Code, and other MCP-compatible clients. It is designed for facts, decisions, preferences, lessons learned, and other context that should survive beyond a single chat or tool.

Start here

• Documentation
• Create an account
• MCP memory server

What we work on

• persistent memory for long-running agents
• retrieval shaped by importance, recency, and feedback
• semantic and relational memory structures
• temporal context, provenance, and conflicting memories
• evaluation and benchmarks for agent memory
• knowledge graphs and controllable levels of abstraction

Research

Semantic Level of Detail for Knowledge Graphs: Discovering Abstraction Boundaries via Spectral Heat Diffusion

Agent Memory: Structure, Time, and Evaluation — a curated research map covering temporal change, structured representations, abstraction, consolidation, and evaluation.

We are building the public artifacts around this work progressively: reproducible examples, datasets, and interactive Spaces.

Status

Mnemoverse is under active development. If you are evaluating memory systems for an agent workflow, the documentation is the best place to begin.