cli-mem 0.3.4

Privacy-first CLI that turns shell history into searchable memory

mem

Your shell history, understood. Not just searched.

A privacy-first CLI that turns your terminal history into an intelligent,
searchable memory system — powered by on-device AI, with zero cloud dependencies.

---

Unlike Ctrl+R, mem knows where you are. The same query returns different results depending on your current git repository — because kubectl apply means something different in your infra repo than in your backend repo.

Unlike cloud-based history tools, nothing ever leaves your machine. Every command, pattern, and session stays in plain text files you can cat, grep, and tail.

Features

• Context-aware search — results ranked by your current git repo, not just recency
• AI pattern extraction — learns that kubectl get pods, kubectl get services, kubectl get deployments are all kubectl get <resource>
• 100% on-device — uses Apple Foundation Models locally. Zero network. Zero telemetry
• Plain text storage — everything in ~/.mem/ as JSONL files. Inspect with cat. Search with grep
• Silent capture — shell hook adds <5ms to prompt. You won't notice it
• Session replay — recall the exact sequence of commands from last Tuesday's debugging session

Quick start

1. Install

# Homebrew (recommended)
brew install matinsaurralde/tap/mem

# Quick install script
curl -fsSL https://raw.githubusercontent.com/matinsaurralde/mem/main/install.sh | bash

# pip / pipx
pipx install mem-cli

2. Activate

echo 'eval "$(mem init zsh)"' >> ~/.zshrc
source ~/.zshrc

3. Use your terminal

Every command is silently captured with full context — directory, git repo, exit code, duration.

4. Search

mem deploy

 1  kubectl apply -f deployment.yaml    infra       2h ago
 2  docker compose up -d                backend     1d ago
 3  fly deploy                          api         3d ago

That's it. mem gets smarter the more you use it.

Usage

Search history

mem kubectl                    # search by keyword
mem "docker compose"           # search by phrase
mem deploy -n 20               # show more results
mem deploy --json              # machine-readable output

See patterns

After running mem sync, mem uses on-device AI to extract structural patterns from your history:

mem kubectl --pattern

Patterns for "kubectl":

  kubectl get <resource>
  kubectl describe <resource> <name>
  kubectl logs <pod> [--tail=<n>]
  kubectl apply -f <file>

Recall sessions

mem session "api outage"

+-----------------------------------------+
| Session: 2026-03-07 14:30  myapp        |
|-----------------------------------------|
|  1  kubectl logs api-7f9b --tail=100    |
|  2  kubectl get pods -n production      |
|  3  kubectl rollout restart deploy api  |
|  4  curl -s localhost:8080/health       |
+-----------------------------------------+

Replay a session? [number/n]: _

More commands

mem stats                      # top commands, repos, totals
mem sync                       # extract patterns + clean old data
mem forget "API_KEY=sk-..."    # permanently delete sensitive commands
mem init zsh                   # print shell hook code

How it works

You type a command
        │
        ▼
   Shell hook (preexec/precmd)
        │
        ▼
   mem _capture  ← runs in background, <5ms
        │
        ▼
   Append one JSON line to ~/.mem/repos/<repo>.jsonl

When you search, mem reads the JSONL file for your current repo and scores each command:

score = (frequency × 0.4) + (recency × 0.4) + (context × 0.2)

• Frequency — how often you've run this exact command
• Recency — exponential decay with a 7-day half-life
• Context — 1.0 if same repo, 0.5 if same directory prefix, 0.0 otherwise

Pattern extraction uses Apple Foundation Models running entirely on-device. No API keys, no cloud calls, no data exfiltration — just your Mac's neural engine.

Storage

All data lives in ~/.mem/ as human-readable plain text:

~/.mem/
  repos/
    infra-k8s.jsonl          # commands from this git repo
    backend.jsonl
    _global.jsonl            # commands outside any repo
  sessions/
    2026-03-05.jsonl         # grouped work sessions
  patterns/
    kubectl.json             # AI-extracted patterns
    docker.json

Every file is inspectable:

cat ~/.mem/repos/myapp.jsonl
tail -f ~/.mem/repos/myapp.jsonl    # watch commands arrive in real-time
grep "docker" ~/.mem/repos/*.jsonl  # search across all repos with grep

Privacy

mem is built on a simple promise: your shell history never leaves your machine.

• Zero network requests — not even update checks
• Zero telemetry — no usage tracking, no analytics, no crash reports
• Zero cloud dependencies — works fully offline, always
• On-device AI only — Apple Foundation Models run on your Mac's neural engine
• Plain text storage — no proprietary formats, no encrypted blobs. You own your data

Read more in PHILOSOPHY.md.

Requirements

Requirement	Version
macOS	26.0+
Python	3.10+
Apple Intelligence	Enabled (for pattern extraction)

Note: mem works without Apple Intelligence — you just won't get AI-extracted patterns. Search, capture, and everything else works fine.

Installation

Homebrew

brew tap matinsaurralde/tap
brew install mem

Quick install

curl -fsSL https://raw.githubusercontent.com/matinsaurralde/mem/main/install.sh | bash

pipx (recommended for Python users)

pipx install mem-cli

From source

git clone https://github.com/matinsaurralde/mem.git
cd mem
pip install -e ".[ai]"     # with AI pattern extraction
pip install -e "."         # without AI (search-only)

Shell setup

After installation, add the hook to your shell:

# zsh (v1)
echo 'eval "$(mem init zsh)"' >> ~/.zshrc
source ~/.zshrc

Bash and fish support coming in v1.5.

Data retention

mem never grows unbounded. Running mem sync automatically cleans old data:

Data	Retention	Rationale
Commands	90 days	High-volume, older ones rarely recalled
Sessions	30 days	Useful for recent postmortems
Patterns	Forever	Small files, accumulated learning

Override defaults: mem sync --keep-commands 180 --keep-sessions 60

Uninstall

brew uninstall mem          # or: pipx uninstall mem-cli
rm -rf ~/.mem               # remove all captured data

Remove the eval "$(mem init zsh)" line from your ~/.zshrc.

Contributing

Contributions are welcome. Please read the PHILOSOPHY.md first to understand the principles that guide this project.

git clone https://github.com/matinsaurralde/mem.git
cd mem
pip install -e ".[dev]"
pytest

License

MIT