chocksy/cems-server repository overview

⁠CEMS

Continuous Evolving Memory System

Persistent memory for AI coding assistants. Your agent remembers what you teach it — across sessions, projects, and teams.

Works with Claude Code · Cursor · Codex · Goose · any MCP-compatible agent

⁠How It Works

1. You code — CEMS hooks into your IDE (Claude Code, Cursor, Codex, Goose)
2. CEMS watches — Every prompt is enriched with relevant memories. Session learnings are extracted automatically.
3. Memories stored — Preferences, conventions, architecture decisions, debugging fixes
4. Context recalled — Next session, relevant memories are injected so your agent already knows your codebase

No manual tagging. No copying context. It just works.

⁠Architecture

⁠Quick Start

⁠1. Deploy the server

# Download the compose file
curl -fsSLO https://raw.githubusercontent.com/chocksy/cems/main/deploy/docker-compose.yml

# Create .env
cat > .env << 'EOF'
POSTGRES_PASSWORD=your_secure_password
OPENROUTER_API_KEY=sk-or-your-key
CEMS_ADMIN_KEY=cems_admin_random_string
EOF

# Start
docker compose up -d

# Run migrations
for f in migrate_docs_schema.sql migrate_soft_delete_feedback.sql migrate_conflicts.sql; do
  curl -fsSL "https://raw.githubusercontent.com/chocksy/cems/main/scripts/$f" | \
    docker exec -i cems-postgres psql -U cems cems
done

⁠2. Create users

cems admin --admin-key $CEMS_ADMIN_KEY users create alice
# Returns the API key — save it, shown only once!

⁠3. Install the client

Each developer runs:

curl -fsSL https://getcems.com/install.sh | bash

Prompts for server URL and API key, then configures your IDE. Done.

⁠Documentation

⁠Development

git clone https://github.com/chocksy/cems.git && cd cems
uv pip install -e ".[dev]"
pytest

⁠License

MIT