mem0-memory

# Mem0 Memory Management

Persist and retrieve semantic memories across Claude sessions.

## Memory Scopes

Organize memories by scope for efficient retrieval:


| Scope                | Purpose                           | Examples                                    |
| -------------------- | --------------------------------- | ------------------------------------------- |
| `project-decisions`  | Architecture and design decisions | "Use PostgreSQL with pgvector for RAG"      |
| `project-patterns`   | Code patterns and conventions     | "Components use kebab-case filenames"       |
| `project-continuity` | Session handoff context           | "Working on auth refactor, PR #123 pending" |

## Project Isolation

Memories are isolated by project name extracted from `CLAUDE_PROJECT_DIR`:

- Project name: `basename($CLAUDE_PROJECT_DIR)` (sanitized to lowercase, dashes)
- Format: `{project-name}-{scope}`

**Edge Case:** If two different repositories have the same directory name, they will share the same `user_id` scope. To avoid this:

1. Use unique directory names for each project
2. Or use `MEM0_ORG_ID` environment variable for additional namespace

**Example:**
- `/Users/alice/my-app` → `my-app-decisions` ✅
- `/Users/bob/my-app` → `my-app-decisions` ⚠️ (collision if same mem0.ai project)
- With `MEM0_ORG_ID=acme`: `/Users/alice/my-app` → `acme-my-app-decisions` ✅

## Memory Categories

Memories are automatically categorized based on content. Available categories:

| Category | Keywords | Use Case |
|----------|----------|----------|
| `pagination` | pagination, cursor, offset | API pagination patterns |
| `security` | security, vulnerability, OWASP | Security patterns and vulnerabilities |
| `authentication` | auth, JWT, OAuth, token | Authentication patterns |
| `testing` | test, pytest, jest, coverage | Testing strategies |
| `deployment` | deploy, CI/CD, Docker, Kubernetes | Deployment patterns |
| `observability` | monitoring, logging, tracing, metrics | Observabil