agent-memory-kit

Agent Memory Kit — Memory Patterns for AI Agents

Memory patterns, context strategies, and persistence recipes for AI agents. The #1 unsolved pain point in production AI agents — session persistence and context management.

Why Agent Memory?

Agent memory is the hottest unsolved problem in AI engineering (2026). Three repos in this space hit 220K+ stars in a single week. Every production agent eventually needs:

• Remembering what happened last turn
• Surfacing relevant past conversations
• Managing context windows without losing information
• Persisting state across sessions
• Choosing what to forget

This kit solves that. Patterns, not abstractions. Grab what works.

What's Inside

Memory Patterns

Pattern	Use Case
Sliding Window	Keep last N turns — simple, cheap, works for short sessions
Summary Compression	Summarize older context to fit within token limits
Vector Recall	Semantic search over past conversations using embeddings
Hybrid Memory	Short-term window + long-term vector recall combined
Episodic vs Semantic	Separate specific events from general knowledge

Memory Strategies

Strategy	Description
Session Persistence	Save/load agent state across restarts
Cross-Session Recall	Remember user preferences and past sessions
Context Windowing	Manage context windows for long-running agents
Cache-Aware Design	Maximize prompt caching for memory operations

MCP Integration

Guide	Description
MCP Memory Server Config	Configure MCP memory servers for agent persistence
MCP Memory Integration	Connect MCP tools to agent memory systems

Memory Prompts

Prompt	Use Case
Memory Injection	System prompt patterns for injecting memory context
Context Summarization	Prompts for summarizing conversation history
Recall Instructions	Instructions for agents to recall and use past info

Framework Guides

Framework	Description
LangGraph Memory	Memory patterns in LangGraph agents
LlamaIndex Memory	Memory in LlamaIndex workflows
CrewAI Memory	Memory in CrewAI multi-agent systems

Quick Start

git clone https://github.com/FreeAutomation-Tech/agent-memory-kit.git
cd agent-memory-kit

# Browse memory patterns
ls patterns/

# Use a pattern
cat patterns/sliding-window.md

# Run an example
python examples/chat-memory.py

# Test memory patterns
python -m pytest tests/

Structure

patterns/            # Core memory patterns
strategies/          # Memory management strategies
mcp/                 # MCP memory server integration
prompts/             # System prompts for memory
frameworks/          # Framework-specific guides
examples/            # Runnable Python examples
tools/               # Memory analysis tools
tests/               # Test files

Usage

# Read a pattern
cat patterns/hybrid-memory.md

# Try the chat memory example
python examples/chat-memory.py

# Analyze your context usage
python tools/context_analyzer.py --log conversation.log

# Test memory patterns
python tools/memory_test.py --pattern sliding-window

Related Repositories

• claude-marketing-kit — Claude Marketing Skills + Python engine (copy, CRO, analytics, email, social, brand)
• claude-seo-kit — Claude SEO Skills + Python engine (audits, schema, sitemap, hreflang, GEO)
• agent-security-kit — Security patterns for AI agents
• claude-prompt-kit — Prompt engineering patterns

License

MIT — see LICENSE.

---

If you find this useful, please consider giving it a star ⭐ — it helps others discover it too!

Thank you for your support! 🙏