Memmy is a personal AI agent and a local memory hub that several AI agents share. It turns local AI collaboration history into structured memory, carries project background, preferences, and progress forward when you switch tools, and can itself organize information and continue unfinished tasks. Agents that can connect include DeepSeek Harness, OpenClaw, Hermes, Claude Code, Codex, Cursor, WorkBuddy, OpenCode, and Pi.
You can use it as a desktop app, as the memmy CLI and TUI, or through the memmy-memory CLI. Signing up in the desktop app grants trial tokens for the hosted memory and agent runtime; when they run out you switch to BYOK mode with your own model API, configured in ~/.memmy/config.yaml with a provider, model, and API key.
On Linux (x64 or arm64, Node.js 22 or later, with a systemd user session), the installer enables a local Memory Service, and the first memmy run opens a model setup wizard, enables a Gateway service, and enters the TUI. Both are systemd --user services bound to localhost and stay up after the terminal closes. Installing does not change Codex, Claude Code, Cursor, or other agents; memmy-memory init (optionally --agent ) installs the Memory Skill and the supported hook or plugin for an agent when you ask for it.
Features
- Shared memory across agents: one local memory store that every connected agent reads and writes, so context survives a switch between tools
- Personal agent: memmy agent --message runs a single-turn task, and the interactive TUI runs multi-turn sessions
- OpenAI-compatible API: memmy serve exposes the agent on port 18990
- Memory service CLI: memmy-memory runs init, health, search, add, and get against the local service on 127.0.0.1:18960, with --url, --token, --source, and --user-id to pick a service and namespace
- Per-agent integrations: a Memory Skill plus hooks or plugins for supported agents, installed only on request
- Roadmap: memory sources beyond AI conversations (browser activity, local documents, devices) and agent-to-agent team collaboration
Integrates with
Claude Code
Terminal-based agentic coding tool that reads your codebase and runs tasks through natural-language commands
Codex
Lightweight coding agent that runs in your terminal
Hermes
Self-improving AI agent with a learning loop that creates and refines skills, recalls past sessions, and runs across chat platforms
OpenClaw
Self-hosted single-operator assistant runtime that links models, tools, and messaging channels through one local Gateway