Guide · Agent skill
How to use the Memory Search agent skill (mem-search)
Updated · written with AI from the project's own documentation and checked against our catalogue data
The mem-search skill teaches an AI agent to look through a persistent memory of earlier work sessions before it answers questions like "did we already solve this?". This guide is for people who have never installed an agent skill and want the agent to reuse past fixes and decisions. By the end you will know what the skill does, how to add it, how to ask for it, and what it needs to work.
Memory Searchthedotmack/claude-mem · Apache-2.0 Download the skill Quick start
- 1Check the requirements. Confirm you have the claude-mem memory system with its MCP search tools, and Node.js with npm if you need to install them.
- 2Download the skill ZIP. Get the mem-search ZIP file from this page.
- 3Unzip into a skills folder. Unzip into ~/.claude/skills/ for personal use, or into the project's .claude/skills/ folder for one project.
- 4Or upload in the Claude apps. In the Claude apps, go to Settings → Capabilities → Skills and upload the skill.
- 5Install claude-mem if tools are missing. If the MCP search tools are not available, run npx claude-mem install --ide <host>.
- 6Ask a natural question. Ask something like "did we already solve this?" so the agent searches past sessions.
- 7Review the results. Check the summaries the agent returns and ask for more detail only when needed.
What the mem-search skill makes the agent do
A skill is a folder with a SKILL.md file that tells an AI agent how to handle a certain kind of task. The mem-search skill covers one task: finding out what was done in earlier sessions. It is built for the claude-mem memory system, which stores records of past work.
The skill asks the agent to search in layers rather than loading everything at once. It first searches for matching entries, then looks at the context around a match, then pulls full details only for the entries that matter. Only as a last resort does it open the original tool inputs and outputs, which can be very long.
- search(query): find matching entries and get their IDs
- timeline(anchor=ID): read nearby context around an entry
- get_observations(ids=[...]): get full details for selected IDs only
- get_tool_uses(ids=[...]): get the original tool input and response for specific calls, used last because these can run to thousands of tokens each
Install the mem-search skill
Download the ZIP file for the skill from this page. Then choose one of two locations. For a personal setup, unzip it into ~/.claude/skills/. For a single project, unzip it into the project's .claude/skills/ folder.
If you use the Claude apps, you can instead upload the skill under Settings → Capabilities → Skills.
- Personal install: unzip into ~/.claude/skills/
- Project install: unzip into the project's .claude/skills/ folder
- Claude apps: Settings → Capabilities → Skills, then upload
Set up the claude-mem memory system
The skill does not work on its own. It depends on the claude-mem memory system and its MCP search tools: search, timeline, get_observations and get_tool_uses. If those tools are missing, the source says to run the install step first, using the npx claude-mem install command with an --ide option for your host.
You need Node.js and npm to run that install command.
npx claude-mem install --ide <host>How to trigger the skill with natural requests
You do not need a special command. The skill is meant to be triggered when you ask the agent to search memory, check whether a problem was already solved, or bring in work from previous sessions. Plain questions work well.
Ask in the way you would ask a colleague. Name the topic so the search has something to match.
- "Did we already solve this CORS error?"
- "How did we set up the staging database last time?"
- "Search memory for the decision about the retry logic."
- "Pull up the work from previous sessions on the login bug."
Safety notes: scripts, network and sensitive projects
According to the skill's listing, mem-search runs no scripts and uses no network. It only searches stored past-session memory, so it cannot fetch fresh information from the internet.
Past tool outputs may contain secrets or private data. If a project is sensitive, do not query its stored session history with this skill. The skill also asks the agent to stamp writes with a platformSource value of cursor or grok-bot for the host, and to avoid dropping the other host when reading unless you ask it to.
- Runs scripts: no
- Uses network: no
- Do not use on projects where stored session history must stay unqueried
What people use it for
- Finding the fix for a bug the team hit in an earlier session
- Recalling a configuration change made weeks ago without rereading old chats
- Checking which decision was made about an architecture choice
- Picking up a task from a previous conversation without starting over
Frequently asked questions
How do I install the mem-search skill?
Download the ZIP from this page and unzip it into ~/.claude/skills/ for personal use or into a project's .claude/skills/ folder. In the Claude apps, you can upload it under Settings → Capabilities → Skills.
What do I need to use mem-search?
You need the claude-mem memory system with its MCP search tools: search, timeline, get_observations and get_tool_uses. If the tools are missing, install claude-mem with npx claude-mem install, which requires Node.js and npm.
Is mem-search free and what is its licence?
The listing states that the skill is licensed under Apache-2.0. The source does not list a price, so check the directory page for any cost details.
Does mem-search search the internet?
No. The skill uses no network and only searches stored past-session memory. It cannot provide fresh information from the internet.
Why does the agent not show full details right away?
The skill uses a layered approach. It searches first, reads nearby context, and only then pulls full details for chosen IDs. Raw tool outputs are a last resort because they are unsummarized and can be very long.
What are alternatives to mem-search?
The source does not compare this skill with other tools, so it does not name alternatives. If you do not use claude-mem, this skill is not suitable because it depends on claude-mem's search tools.
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 mem-search if you work in Cursor or a similar agent with claude-mem and want to reuse past fixes, decisions and configuration work. Skip it if you have not installed claude-mem, need current information from the internet, or work on sensitive projects where stored session history should not be queried.
Download the skill