claude-mem vs Memlo: which memory to use with Claude Code
Last updated
claude-mem records what your coding agent does through hooks, summarizes it with a model, and keeps it on your own computer. Memlo is a hosted memory server: agents save the memories they choose, and every agent you sign in to, on any computer, searches the same ones. Pick claude-mem if you want free, local, automatic capture for your own sessions. Pick Memlo if you want one curated memory shared across agents, computers and people.
At a glance
Facts about claude-mem come from its own documentation, read on October 30, 2026. Links are in Sources at the end.
| claude-mem | Memlo | |
|---|---|---|
| What it is | An open-source plugin that captures sessions | A hosted MCP memory server |
| How memories are made | Automatic: hooks record tool use and a model summarizes it | The agent saves a memory with a tool when something is worth keeping |
| Where data lives | On your computer (~/.claude-mem/), in SQLite with optional Chroma; optional paid cloud sync | In your Memlo account, hosted in the United States |
| Install | Claude Code plugin or npx claude-mem install; needs Node.js 20+ | Nothing to install: add https://memlo.ai/mcp and sign in |
| Agents | Claude Code first, with installers for Codex CLI, Cursor, Windsurf, OpenCode and others | Claude Code, Codex, claude.ai, Claude Desktop, ChatGPT, Gemini, VS Code and other MCP clients that sign in with OAuth |
| Model for summaries | Your own Anthropic plan, another provider's key, or the paid CMEM observer | Included; it counts toward a daily limit per person |
| History of a memory | Not documented | Every change is a version you can view and restore |
| Sharing | TeamBrain, a pilot for teams, with custom pricing | Share a space or a page by email, as Viewer or Editor, or with a link |
| License | Apache 2.0 | Not open source |
| Price | Free and open source; CMEM Pro from $30 a month, with a 30-day trial | Free during the invitation-only beta |
| Can you start today | Yes | By invitation; join the waitlist |
How claude-mem works
claude-mem installs as a Claude Code plugin. Hooks run at the start and end of a session, when you send a prompt and around each tool call. They record what the agent did, and a background worker uses a model to compress those records into short summaries. The next session starts with relevant summaries injected into its context.
The agent finds older work with claude-mem's search tools in three steps: a short index of matches, a timeline around one, then the full records it picked. The documented MCP tools are for searching. Memories come from the automatic capture, not from the agent choosing to save.
Everything stays on your computer unless you turn on CMEM Pro's cloud sync.
How Memlo works
Memlo is an MCP server at https://memlo.ai/mcp. You add it to each agent and sign in once. The agent then has tools to save, search, read, change and retire memories. It saves a memory when it learns something worth keeping, such as a decision and why it was made, and it searches before it acts.
- One memory for every agent. Claude Code, Codex, claude.ai and ChatGPT find the same memories when they sign in to the same account.
- History. Every change is kept as a version; an outdated memory can be retired without deleting it.
- Spaces and sharing. Keep memories in spaces, and share a space or a single page with other people.
Where claude-mem is the better choice
- You want nothing to leave your computer. claude-mem stores everything locally by default. Memlo is hosted.
- You don't want to think about saving. claude-mem captures every session on its own. With Memlo, the agent decides what to save, which keeps memories short but can miss something.
- You want it now, and free. claude-mem is open source and anyone can install it today. Memlo is by invitation.
- You want to read or change the code. claude-mem is Apache 2.0.
Where Memlo is the better choice
- You use more than one agent or computer. The same memories reach Claude Code, Codex and chat apps, wherever you sign in.
- You want curated memories, not session logs. Each memory is a short note with a title, so search returns decisions and facts rather than every step of past work.
- You want to see how a fact changed. Every memory keeps its versions.
- You share knowledge with other people. Spaces and pages can be shared by email with a role.
- You don't want to install or run anything. There is no package, worker or database on your computer, and no API key.
Can you use both?
Yes. They don't conflict: claude-mem's hooks capture sessions locally while Memlo holds the decisions you want every agent to share. If you run both, give each a clear job so the agent knows where to look.
Frequently asked questions
Is claude-mem free?
The open-source plugin is free; you pay only for the model that writes its summaries, which can be your own Anthropic plan. CMEM Pro, with a hosted summarizer, cloud sync and remote recall, starts at $30 a month after a 30-day trial (as listed on October 30, 2026).
Does claude-mem work with Codex?
Its installation guide lists Codex CLI among the supported agents, along with Cursor, Windsurf and others.
Does Memlo capture sessions automatically?
No. Agents save memories with Memlo's tools when something is worth keeping. A SessionStart hook in Claude Code makes each session begin by loading your key memories.
Where does claude-mem store data?
In ~/.claude-mem/ on your computer, in a SQLite database, with an optional Chroma database for meaning-based search.
Sources
Checked on October 30, 2026:
- claude-mem on GitHub and its license
- claude-mem installation
- claude-mem search tools
- claude-mem cloud sync and TeamBrain
- CMEM pricing
- Memlo: the setup page and terms
Try Memlo
Memlo is invitation only for now. Join the waitlist and we'll email you when your invitation is ready.
Join the waitlist