Memmy logo

Memmy

Local-first personal memory hub plus agent runtime shared across Cursor, Claude Code, Codex, and more.

Open source

Memmy (MemTensor/memmy-agent, MIT, v1.0.3) is a local-first personal memory hub and agent runtime: it imports history from Cursor, Claude Code, Codex, OpenCode, OpenClaw, and Hermes Agent into a MemOS-powered memory layer, exposes that memory to external agents via Skills/MCP and a memmy-memory CLI, and runs tasks through a desktop app, CLI/TUI, or OpenAI-compatible API. Signup grants 30 million trial tokens; afterward use BYOK with your own providers while keeping memory and app state on the machine by default.

Selection advice

Choose Memmy when several coding agents must share one inspectable personal memory and you want a local runtime entry point. Choose Emulo when you mainly need a mined working profile from logs; choose Mem0 when memory lives inside application code you ship.

Quick comparison

Memmy combines a cross-agent memory layer with a local agent runtime and history onboarding. Emulo mines session logs into a working profile; Mem0 embeds memory mainly through application SDKs.

MemmyEmuloMem0
Best forOne shared personal memory across multiple coding agents plus a local runtimeMining coding-agent transcripts into a portable you.md-style profileEmbedding user memory inside a product you ship
Primary surfaceDesktop app, CLI/TUI, OpenAI-compatible API, memmy-memory CLI, Skills/MCPLocal mining engine + plugins/skills/MCP profile loaderSDK/API memory with optional hosted platform
TradeoffFull local hub/runtime vs more to install and configure than a drop-in SDKStrong profile mining vs less of a full multi-entry agent runtimeProduct-embedding flexibility vs weaker cross-harness desktop memory UX

Best for

  • cross-agent shared personal memory
  • importing existing Cursor/Claude Code/Codex history
  • local-first desktop + CLI agent runtime
  • BYOK after trial tokens

Not ideal for

  • teams that only need an SDK memory library inside their own product
  • fully managed cloud-only memory with no local install
  • Linux-first desktop workflows (docs list macOS and Windows downloads)

Core concepts

MemOS-powered memory enginecross-agent shared memoryagent history onboardinglocal agent runtimeBYOK / Account mode trial tokensSkills and MCP extensions

Minimal implementation shape

Install the desktop app or run `memmy onboard`, then `memmy-memory init` / `memmy-memory search "…"` to wire Skills for external agents; or `memmy serve` for the OpenAI-compatible API on :18990.

Sources