Memex Cli
Memex, Claude memory via local RAG (MCP, offline embeddings)
Last verified:
What is Memex Cli?
memex is a local memory layer for Claude that automatically saves useful insights from conversations and retrieves them semantically when needed. It is positioned as a “second brain” for Claude, helping the model remember preferences, decisions, and context across sessions. The site emphasizes that memory stays on the user’s machine, with no cloud dependency and no API keys required. It is designed for people who use Claude regularly and want persistent, private memory without manually copying context between chats.
The product is delivered as a CLI that you install globally, then register as an MCP server so it can work with Claude Code and Claude Desktop. On first run, it downloads a local embedding model of about 450MB into the user’s home directory. The website highlights that the memory is surfaced semantically, meaning it tries to retrieve the most relevant past information rather than relying only on exact keywords. That makes it useful for long-running projects, recurring workflows, and personal context that should carry forward.
memex is aimed at developers and power users who want Claude to behave more consistently across conversations. It is especially relevant for people who care about local-first storage, privacy, and integrating with the MCP ecosystem. The experience appears intentionally lightweight and terminal-driven, which fits users comfortable with CLI tools and AI coding workflows. It is less about generic note-taking and more about giving Claude a durable working memory.
The site also suggests that the tool is meant to reduce repetitive explanation and improve continuity in multi-session work. By automatically storing insights and recalling them later, it helps keep preferences, project details, and prior decisions available without retyping them. Because it is local, it may appeal to users who want more control over their data. Its main value is making Claude feel stateful instead of session-bound.
Memex Cli pricing
Pricing model: Freemium
The website presents memex as free to install from npm with no cloud account, no API keys, and no hosted plan described. The visible setup instructions say to install it globally with npm and note that the embedding model is downloaded locally on first run. No paid tiers, subscription pricing, or premium feature bundles are shown on the website, so the listed offering appears to be a free local CLI tool.
Memex Cli pros
- Local-first memory storage
- No cloud required
- No API keys needed
- Automatic insight saving
- Semantic retrieval of past context
- Persistent memory across conversations
- Works with Claude Code
- Works with Claude Desktop
- MCP server support
- Simple CLI installation
- Global npm install workflow
- Embedding model downloaded once
- Memory stays on the user’s machine
- Reduces repeated context setup
- Useful as a Claude second brain
- Designed for long-running projects
- Privacy-focused architecture
- Developer-friendly integration
Memex Cli cons
- Requires command-line comfort
- Depends on Claude ecosystem
- Needs local model download
- Initial download is large
- Not a full general-purpose note app
- Best value is tied to Claude usage
- Setup includes MCP registration
- Local storage may not suit shared teams
- No cloud sync mentioned
- Not ideal for non-technical users
- Only stores data on one machine
- Memory quality depends on embeddings
- Limited usefulness outside AI workflows
- No mobile experience mentioned
Frequently asked questions about Memex Cli
What does memex do?
memex gives Claude persistent memory across conversations by automatically saving useful insights and surfacing them later when they are relevant. It is meant to make Claude feel like it remembers you, your preferences, and your ongoing work instead of starting from scratch each time.
Is memex cloud-based?
No. The website says all data stays on your machine, and it explicitly says there is no cloud. That local-first design is part of its privacy story and means your memory is stored and processed locally rather than in a hosted service.
Do I need API keys to use it?
No API keys are required for memex itself according to the site. The product is positioned as working locally without cloud services, and the installation instructions do not ask for an account or hosted credentials.
How do you install memex?
The website shows a global npm install command for the CLI. After installation, you register memex as an MCP server so it can connect with Claude Code and Claude Desktop.
What AI tools does it work with?
The site says memex works with both Claude Code and Claude Desktop. Its MCP server setup is what enables that integration, so it is specifically aimed at Claude-based workflows rather than being a general assistant memory layer for every AI tool.
What happens on first run?
On first run, memex downloads its embedding model locally, and the website notes that the download is about 450MB. It stores that model under the user’s home directory in ~/.memex/models/.
Why is semantic memory important here?
Semantic retrieval lets memex find relevant past insights even when the wording is different from the original conversation. That is useful for memory because users often remember concepts or decisions, not exact phrases.
Who is memex for?
It is aimed at Claude users, especially developers and power users who want persistent memory across sessions. The CLI and MCP-based setup make it a strong fit for people already working in terminal-driven or AI coding environments.
Does memex store data online anywhere?
The website says no, and it emphasizes that all data stays on your machine. That means the tool is designed to be private and local rather than synced through a hosted memex account.
What is memex trying to solve?
memex is trying to solve the problem of session-bound AI context. Instead of re-explaining preferences, project details, and prior decisions in every conversation, it saves those insights and makes them available later so Claude can stay consistent over time.