Documentation

Agent-Memo.AI gives your AI coding agent persistent cloud memory. Decisions, preferences, architecture choices, and project context are stored once and recalled across every session.

Setup Guide

Get started in 2 minutes. Add two config files and your agent has persistent memory.

MCP Integration

14 tools via Model Context Protocol. Works with Claude Code, Cursor, Windsurf, and any MCP client.

REST API

Full HTTP API for memory CRUD, semantic search, topic management, and knowledge graph operations.

Knowledge Graph

Store entity relationships and query connected facts. Track how decisions and architecture evolve over time.

Quick start

# 1. Sign up and get your API token

open https://agent-memo.ai/register

# 2. Add .mcp.json + .claude/settings.local.json to your project

See setup guide → agent-memo.ai/docs/setup

# 3. Start Claude Code — your agent now has memory

claude

Key concepts

Memories: Discrete units of knowledge — decisions, preferences, facts, feedback. Each memory has a type, topic, importance level, and optional tags.
Projects: Memories are scoped to projects. Each repository or workspace gets its own isolated memory space. Auto-detected from git remote or directory name.
Knowledge Graph: Entity-relationship store for structured facts. Track how components connect, who owns what, and how architecture decisions relate.
Semantic Search: BGE-M3 embeddings enable cross-language retrieval. Find memories by meaning, not just keywords.