← memex

Download memex

v1.0.0

Apache 2.0 licensed. ~50 MB Tier 0, ~80 MB with vector retrieval. Ships on PyPI as memex and on Docker Hub as quefly/memex.

View source on GitHub Release notes ↗ PyPI ↗

pipx (recommended)

macOS · Linux · Windows
pipx install memex

Isolated install — keeps memex out of your system Python. Requires Python 3.10+.

uv

macOS · Linux · Windows
uv tool install memex

Same isolation as pipx, faster resolver. Requires uv installed.

pip (in a venv)

macOS · Linux · Windows
pip install memex

For when you want memex inside a project venv rather than globally.

Docker

Any OS with Docker
docker run -d --name memex \
  -v $HOME/.memex:/data \
  -p 127.0.0.1:7777:7777 \
  quefly/memex:1.0.0

Multi-arch images (amd64, arm64). Signed with cosign keyless via GitHub OIDC.

With vector retrieval

macOS · Linux · Windows
pipx install --force 'memex[embed]==1.0.0'

Adds fastembed (~30 MB model) for semantic similarity recall on top of BM25.

From source

macOS · Linux · Windows
git clone --branch v1.0.0 https://github.com/queflyhq/memex.git
cd memex
uv sync --all-extras

For contributors. See the contributing guide for the full dev loop.

Verify the Docker image

Images are signed with cosign keyless via GitHub OIDC. No public keys to manage — verify against the GitHub identity that built the image.

cosign verify quefly/memex:1.0.0 \
  --certificate-identity-regexp "https://github.com/queflyhq/memex/.+" \
  --certificate-oidc-issuer https://token.actions.githubusercontent.com

After install

  1. Sanity check: memex doctor
  2. Bootstrap the meta-skills: memex install skill:using-memex and memex install skill:core-validations
  3. Wire into Claude Code: claude mcp add memex memex serve
  4. Read the quickstart for the full 60-second walkthrough.