CLI Reference

The memories CLI is the primary interface for managing your AI agent memory store. All commands operate on a local SQLite database at ~/.config/memories/local.db.

Command references on this page map to a segmented lifecycle model: session working memory + long-term semantic/episodic/procedural stores. See Memory Segmentation.

Quick Reference

# First-time setup (auto-configures MCP for detected tools)
memories setup

# Add memories
memories add --rule "Always use TypeScript strict mode"
memories add --decision "Chose Tailwind for utility-first CSS"
memories add --fact "API rate limit is 100/min"

# Search
memories search "authentication"
memories search "how to handle login" --semantic

# Reminders
memories reminders add "0 9 * * 1-5" "Review open TODOs"
memories reminders run

# Generate instruction files
memories generate all

# Sync config files
memories files ingest
memories files apply --global

Command Overview

Setup Commands

Command	Description
`init`	Initialize memories and auto-configure MCP for detected tools (`setup` alias)
`login`	Authenticate for cloud sync features

Core Commands

Command	Description
`add`	Add a new memory (rule, decision, fact, or note)
`recall`	Get rules and relevant memories for AI agents
`prompt`	Output memories formatted for AI system prompts

Search Commands

Command	Description
`search`	Keyword or semantic search across memories
`list`	List memories with optional filters
`embed`	Generate embeddings for semantic search

Management Commands

Command	Description
`edit`	Edit an existing memory
`forget`	Soft-delete memories by ID or bulk filter
`tag`	Bulk tag and untag operations
`reminders`	Manage cron-based reminders (`add`, `list`, `run`, `enable`, `disable`, `delete`)
`stale`	Find stale memories with superseded/conflict filters
`review`	Interactively keep/delete stale memories

Lifecycle Commands

Command	Description
`session`	Manage explicit memory session lifecycle (`start`, `checkpoint`, `status`, `end`, `snapshot`)
`compact`	Run inactivity compaction worker for active sessions
`consolidate`	Merge duplicates and supersede outdated memories
`openclaw memory`	Read/write OpenClaw semantic, daily, and snapshot files

Generation Commands

Command	Description
`generate`	Generate IDE rule files from memories
`diff`	Show what changed since last generate

Data Commands

Command	Description
`export`	Export memories to JSON or YAML
`import`	Import memories from JSON or YAML
`ingest`	Import from existing IDE rule files

Sync Commands

Command	Description
`sync`	Manage cloud sync
`files`	Sync config files (.agents, .cursor, .claude, etc.)

System Commands

Command	Description
`config`	Manage agent configuration
`org`	List and switch active organization workspace
`serve`	Start the MCP server
`stats`	Show memory statistics
`doctor`	Check health and diagnose issues
`hook`	Manage git hooks

Global Behavior

Project Detection

When inside a git repository, commands automatically scope to the current project (detected from the origin remote URL)
When outside a git repo, or when --global is passed, memories are stored with global scope
Global memories are included in all project queries

Database

All memories are stored in a single SQLite database at ~/.config/memories/local.db
FTS5 full-text search with BM25 ranking
Vector embeddings for semantic search (dimension depends on selected model; default is 384)
Soft-deleted memories are retained but excluded from queries

MCP Integration

When you run memories setup (or memories init), it automatically:

Detects installed AI tools (Cursor, Claude Code, Windsurf, VS Code)
Configures MCP for each detected tool
Generates instruction files with your existing memories
Imports existing project skill configs (such as .claude/skills and .codex/skills) into memories

This means your memories are immediately available to your AI tools via MCP.

On this page