vmem — Memory engine for AI agents

Memory your agents can actually use.

Store context from chats and tools, connect it in a graph, and retrieve the slice that matters — with a Context Trace that shows why each memory matched.

How it works

Store — Capture context from chat, tools, and HTTP — tagged by profile, so workspaces stay separate.
Connect — Memories link in a graph. Relationships survive between sessions instead of sitting in a flat list.
Recall — Agents pull the right slice via MCP, skills, or the API — and get a Context Trace with the answer.

Features

Graph-native memory — See context as a network, not a flat list.
Built for recall — Episodic, knowledge, and profile memories that persist.
Agent-ready — MCP, HTTP API, and skills your agents can call.

Surfaces

MCP, HTTP, and an SDK on the same graph.

Get started with vmem