Overview
Memmy is a local-first memory hub for people who work across multiple AI tools. It scans authorized local histories from tools such as Cursor, Claude Code, Codex, OpenCode, OpenClaw, and Hermes, then distills conversations, preferences, project context, decisions, and pitfalls into structured memory.
The product solves a common problem in modern AI agent workflows: each tool knows only what happened in its own session. Memmy creates a shared memory layer so switching from Cursor to Claude Code or Codex does not mean starting context from scratch.
Memmy ships as a desktop app, CLI, and API. The official site positions it as free and open source for macOS and Windows users, with memory stored locally and no automatic collection without user authorization.
Key Features
Cross-agent memory scanning - Reads authorized local histories from Cursor, Claude Code, Codex, and other agents, then converts scattered conversations into structured memory.
Local-first storage - Memmy's docs describe local SQLite-backed memory, giving developers a stronger privacy posture than cloud-only memory tools.
Context injection across tools - When users switch agents, Memmy can hand over relevant preferences, project background, and recent decisions rather than dumping entire transcripts.
Desktop, CLI, and API access - The desktop app works for everyday interaction, while CLI and API entry points let advanced users connect scripts or custom agents.
Agent runtime capabilities - Beyond memory, Memmy includes task workbench, sessions, tool calls, MCP, scheduled tasks, and skills for more complete AI productivity workflows.
Open-source community path - The project links to GitHub and Discord, making it inspectable for developers who want to understand or modify the memory layer.
Pricing & Plans
Memmy is currently positioned as free and open source. The official homepage lists macOS and Windows support and describes the product as "Free & open source."
| Plan | Price | Includes |
|---|---|---|
| Memmy desktop / CLI | Free | Desktop app, local memory, agent history scanning, CLI/API access, community source |
The docs also mention account mode with trial tokens and API-key mode. For ongoing model usage, users should expect to bring their own model quota or API keys when needed.
Best For
- Developers who regularly switch between Cursor, Claude Code, Codex, OpenCode, or OpenClaw
- Solo builders who want a local memory layer for long-running repo work
- Teams evaluating persistent AI coding context without committing to a cloud memory product
- Power users who want CLI/API access to their own agent memory
- Privacy-conscious users who prefer local storage over opaque hosted memory
FAQ
What is Memmy?
Memmy is a local-first memory hub and agent runtime. It turns authorized AI work history into structured memory that can be reused across tools such as Cursor, Claude Code, Codex, and OpenClaw.
Is Memmy free?
Yes. The official homepage describes Memmy as free and open source for macOS and Windows. Users may still need their own model API keys or model quota depending on how they run it.
Where does Memmy store memory?
Memmy's docs describe a local memory service backed by SQLite by default. The homepage also states that memory data stays on the user's own device.
Does Memmy automatically read every AI conversation?
No. The official site says history scanning requires explicit user authorization and can be revoked. Users decide which agents can use which memory.
How is Memmy different from a notes app?
A notes app stores manually written summaries. Memmy is designed to scan agent histories, structure preferences and decisions, retrieve relevant context, and inject that memory back into agents.
Does Memmy work only with coding tools?
Coding-agent memory is the clearest use case, but Memmy also includes a desktop workbench, channels, tools, sessions, scheduled tasks, and API access that can support broader agent workflows.
Who should skip Memmy?
Skip it if you use only one AI tool, do not want to run local services, or need a mature enterprise memory platform with centrally managed compliance controls.
