Guide · Agent skill
How to use the MemOS Local Memory Installer agent skill
Updated · written with AI from the project's own documentation and checked against our catalogue data
The memos-local skill is an instruction package for an AI agent. It tells the agent how to install, configure, upgrade and troubleshoot MemOS, a memory plugin for the OpenClaw agent platform. After reading this guide you will know what the skill does, how to add it to Claude, how to ask for it in plain language, and what it changes on your machine before you approve it.
Quick start
- 1Confirm OpenClaw is installed. The skill only applies to the OpenClaw agent platform. If you do not run OpenClaw, stop here.
- 2Read the SKILL.md file. Review the instructions, especially the permissions section, before you allow script execution.
- 3Add the skill folder. Download the ZIP and unzip it into ~/.claude/skills/, or into .claude/skills/ in a project, or upload it under Settings → Capabilities → Skills in the Claude apps.
- 4Ask for the install in plain words. Start a new conversation and say something like "install MemOS" or "setup memory".
- 5Choose the embedding model. When asked, pick the default local model for fully on-device memory, or an external API if you have a key and prefer it.
- 6Check the handoff. After installation, read the handoff message and decide whether to accept the optional summarizer or team sharing settings.
What an agent skill is and what this one does
An agent skill is a folder containing a SKILL.md file. The file gives an AI agent instructions for a specific task and tells the agent when to use them. This skill's SKILL.md is the only file listed for it.
Once MemOS is set up, the plugin automatically saves conversations to a local SQLite database and brings back relevant past context in later sessions, so the agent does not start from zero each time. The skill's job is to get that plugin installed and working. It also covers upgrading the plugin and fixing problems such as memory not working or plugin errors.
MemOS also includes a local web Memory Viewer at http://127.0.0.1:18799 for browsing stored memories, tasks and skills. Everyday search and recall are handled by a separate bundled skill, memos-memory-guide, not by this one.
- Stores all data in ~/.openclaw/memos-local/memos.db on your machine
- Requires no API keys, cloud accounts or registration to start
- Uses a local offline embedding model by default; an external API is optional
Install the memos-local skill in Claude
Option 1: download the skill ZIP from the page where you found this guide and unzip it into your personal skills folder, ~/.claude/skills/. To limit it to one project, unzip it into that project's .claude/skills/ folder instead.
Option 2: in the Claude apps, open Settings, go to Capabilities, and upload the skill under Skills.
Start a new conversation after adding the skill so the agent loads it.
- Personal: ~/.claude/skills/
- Project only: .claude/skills/ inside the project folder
- Claude apps: Settings → Capabilities → Skills
How to trigger the skill with natural requests
You do not need a special command. Describe what you want in ordinary words. The skill lists these trigger phrases in its description, and the agent matches requests against them.
Phrases that start installation or setup include "install memos", "install MemOS", "add memory plugin", "setup memory" and "enable memory". For problems, the skill responds to "memory not working", "memories lost" and "memory plugin error". For maintenance, it responds to "upgrade MemOS" and "update memory plugin". It also covers configuration requests such as "configure embedding" and "configure summarizer".
- "Install MemOS on this machine."
- "My memory plugin error says the gateway is not loading it. Can you fix it?"
- "Upgrade MemOS to the latest version."
- "Configure the embedding model to use a local default."
What the skill asks you and what it does on its own
The skill is written so the agent runs the installation without stopping for approval. Before a fresh install, the agent asks one question: whether to use the default local embedding model or an external API. After installation, it gives a handoff message and offers optional settings such as a summarizer or team sharing, then waits for your choice.
The skill states that by loading it you authorize the agent to run shell commands and scripts, download and run install scripts, install or upgrade Node.js and npm packages, edit files under ~/.openclaw/, modify ~/.openclaw/openclaw.json, and start, stop and restart the OpenClaw gateway.
- The only pre-install question is the embedding model choice
- Steps are meant to run one at a time, each checked before the next
- The agent is told to check the system state rather than ask you which path to take
Safety notes before you enable it
This skill runs scripts and uses the network. It tells the agent to download and execute remote scripts, using the forms curl ... | bash and irm ... | iex, as a fallback, and to edit ~/.openclaw/openclaw.json without asking. These actions can change system state in ways you may not review.
Read the SKILL.md file before you install it. If you do not want unattended script execution, do not enable this skill, or install MemOS yourself by following the steps you have reviewed. Only use it on a machine where you accept those changes.
- Runs scripts: yes
- Uses network: yes
- Changes files under ~/.openclaw/, including openclaw.json
- Installs or upgrades Node.js and npm if needed
Tips for a smooth setup
Confirm you run OpenClaw first. The skill is written for that platform and its configuration files, and installing it on a machine without OpenClaw adds system changes with no benefit.
If you want memory to stay fully offline, keep the default local embedding model. Choosing an external provider means you will need that provider's API key.
After setup, use the Memory Viewer to check what has been stored. Because data lives in a single SQLite file, you can also back up or inspect ~/.openclaw/memos-local/memos.db directly.
- Check OpenClaw is installed before you start
- Keep the default local embedding model for offline use
- Use the Memory Viewer to verify what is saved
What people use it for
- Setting up persistent memory so an OpenClaw agent remembers earlier project decisions across sessions
- Keeping all conversation history on your own machine with no cloud account
- Upgrading an existing MemOS plugin when you are asked to update the memory plugin
- Diagnosing a memory plugin error or memories that seem to be missing after a restart
- Browsing stored memories, tasks and skills in the local Memory Viewer
Frequently asked questions
Is MemOS free to use?
The source describes MemOS as Apache-2.0 licensed and says it needs no API keys, cloud accounts or registration to get started. The skill's own licence is listed as Apache-2.0 on the directory page. Check the repository licence file for full terms.
What do I need to run the memos-local skill?
You need the OpenClaw agent platform, Node.js and npm (the install may install or upgrade them), and shell access on your machine. An external embedding or summarizer API key is optional and only needed if you choose a remote provider.
Does MemOS send my data to the cloud?
The source says MemOS stores everything locally in ~/.openclaw/memos-local/memos.db and uses a local offline embedding model by default. Choosing an external embedding or summarizer provider would send the relevant data to that provider.
How do I use the MemOS Memory Viewer?
After the plugin is running, open http://127.0.0.1:18799 in your browser. The viewer lets you browse and manage memories, tasks and skills.
Why is the skill asking me to run curl or irm commands?
The skill uses curl ... | bash or irm ... | iex as a fallback to download and run install scripts. Read the SKILL.md before allowing this, because the skill tells the agent to run these without asking.
Is there an alternative if I do not use OpenClaw?
The source does not name alternatives. This skill is written only for OpenClaw and its configuration files, so it is not a fit for other agent platforms.
New agent skills, repos and AI tools — every week
The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.
Our verdict
Use this skill if you run OpenClaw agents and want persistent memory stored on your own machine, or if you already have MemOS and need help upgrading or fixing it. Pick something else if you do not use OpenClaw, if you do not want an agent running install scripts and editing your configuration without asking, or if you only need memory for a short one-off task.
Download the skill