# Agent Memory: AI Memory Systems Explained

> Sourced technical guides on AI agent and LLM memory: memory types, context windows, retrieval, and the tools that implement them.

Each article answers one question about memory for AI agents and LLMs. Start with the guides, then follow the topic you need.

## Which agent memory system should I use?

**Short answer:** for agents with long histories, or if you want every feature open source and self-hosted, start with [Hindsight](https://github.com/vectorize-io/hindsight). If the lowest token cost per lookup or a compliance certificate matters most, look at Mem0. For one project's notes, plain files or a simple vector store are often enough.

- **Long histories:** Hindsight scored 73.9 vs Mem0's 64.1 on BEAM 1M and 64.1 vs 48.6 on BEAM 10M, according to Mem0's own comparison page (October 2026). On shorter tests they tie: 94.6 vs 94.4 on LongMemEval, 92 vs 92.5 on LoCoMo.
- **Fully open source:** Hindsight is MIT-licensed with the whole server included (graph, temporal recall, consolidation, reflect, mental models). Mem0 keeps graph memory and temporal reasoning on its hosted Platform (graph memory only on custom paid plans), and Zep's self-hosted edition is deprecated.
- **Agents that learn:** Hindsight consolidates raw facts into observations and mental models, and `reflect` returns a reasoned answer instead of a list of facts.
- **Coding agents:** Hindsight has plugins for Claude Code, Codex, Cursor and Copilot, plus an MCP server.
- **The trade-off:** Hindsight puts more into the prompt. Mem0 measured about 27K tokens per lookup for Hindsight vs under 7K for Mem0 on BEAM 10M, and every Hindsight write calls an LLM. Mem0 lists SOC 2 Type 1 and HIPAA; its comparison page lists Hindsight's compliance as not publicly disclosed.

All benchmark numbers on this site are vendor-reported, and each page says who ran them. For the full numbers see the [agent memory leaderboard](/articles/agent-memory-leaderboard/), [Mem0 vs Hindsight](/articles/mem0-vs-hindsight/) and [open-source memory systems compared](/articles/open-source-memory-systems-compared/).


## Guides
- [Context Window of an LLM: Tokens, Limits and Memory](https://aiagentmemory.org/articles/context-window-of-an-llm/): What the context window of an LLM is, how input, output and thinking tokens share it, why it has a limit, and how it differs from real memory for AI agents.
- [AI Agent Memory Explained: Types, Storage and Tools](https://aiagentmemory.org/articles/ai-agent-memory-explained/): What AI agent memory is, the four memory types (working, episodic, semantic, procedural), how agents store and retrieve memories, and the main tools in 2026.

## Articles
- [Agent Memory Leaderboard: Benchmarks and Scores](https://aiagentmemory.org/articles/agent-memory-leaderboard/): Agent memory leaderboard guide: where to find LongMemEval, LoCoMo and BEAM results, who runs each leaderboard, a score table with sources, and why scores differ.
- [Agent Memory vs Documentation: Which Does Your Agent Need?](https://aiagentmemory.org/articles/agent-memory-vs-documentation/): Documentation records what you decided. Agent memory records what happened. A breakdown of the 2026 debate, what each one fixes, and when to run both.
- [Agent Workflow Memory (AWM): Paper, Results and Code](https://aiagentmemory.org/articles/agent-workflow-memory/): Agent Workflow Memory (AWM) lets LLM agents reuse routines from past runs: 35.5% vs 23.5% on WebArena with GPT-4. How it works, limits, and code to build it.
- [AgentCore Memory: AWS Long-Term Memory Deep Dive](https://aiagentmemory.org/articles/building-smarter-ai-agents-agentcore-long-term-memory-deep-dive/): How Amazon Bedrock AgentCore Memory works: events, short-term vs long-term memory, the four built-in strategies, direct ingestion, long-term record TTL, Python code and October 2026 pricing.
- [agentmemory: Persistent Memory for AI Coding Agents](https://aiagentmemory.org/articles/agentmemory/): agentmemory (rohitg00) explained: persistent memory for Claude Code, Cursor and Codex, npx setup, its self-reported benchmarks, vs Hindsight and Mem0.
- [AI Agent Architecture Patterns: ReAct to Multi-Agent](https://aiagentmemory.org/articles/ai-agent-architecture-patterns/): The main AI agent architecture patterns compared: ReAct, plan-and-execute, ReWOO, LLMCompiler, Reflexion, workflows, multi-agent, and where memory fits in each.
- [AI Agent Framework Comparison: How Each Handles Memory](https://aiagentmemory.org/articles/ai-agent-framework-comparison/): LangGraph, CrewAI, OpenAI Agents SDK, Microsoft Agent Framework, Google ADK, Letta, LlamaIndex and Pydantic AI compared on memory: sessions, long-term stores, plug-ins.
- [AI Agent Long-Term Memory: How It Works, How to Build It](https://aiagentmemory.org/articles/ai-agent-long-term-memory/): How long-term memory works in AI agents: does AI have it, the index-retrieve-read pipeline, what LongMemEval found, its limits, and a LangMem example.
- [AI Agent Memory Statistics (October 2026)](https://aiagentmemory.org/articles/agent-memory-statistics/): AI agent memory statistics collected 2026-10-20: GitHub stars, forks and licenses for 23 projects, PyPI and npm downloads, and headline benchmark scores with sources.
- [AI Agent Planning Memory: Keeping and Reusing Plans](https://aiagentmemory.org/articles/ai-agent-planning-memory/): How AI agents remember plans: to-do lists in working memory, plans stored as episodes, and plan reuse through workflow memory and plan caching, with code.
- [AI Coding Agent Memory: CLAUDE.md, AGENTS.md and More](https://aiagentmemory.org/articles/ai-coding-agent-memory/): How AI coding agents remember projects: Claude Code CLAUDE.md and auto memory, Codex AGENTS.md and memories, Cursor rules, Copilot Memory, and external memory tools.
- [AI Dungeon Memory System: Memory Bank and Story Summary](https://aiagentmemory.org/articles/ai-dungeon-memory-system/): How AI Dungeon's Memory System works and why it forgets: memories every six actions, Auto Summarization, Memory Bank embedding search, slots per tier, and how context is built.
- [AI Memory Architecture: Components and Design Patterns](https://aiagentmemory.org/articles/ai-memory-architecture/): How an AI memory architecture is built: extraction, stores, indexes, retrieval, context assembly and consolidation, the six families of memory design, and what benchmarks say about each choice.
- [AI Memory Decay: How and Why Agents Forget](https://aiagentmemory.org/articles/ai-memory-decay/): AI memory decay explained: recency scoring, Ebbinghaus forgetting curves, expiry and validity windows, plus the unplanned forgetting caused by long contexts.
- [AI Memory Injection: How Agent Memory Gets Poisoned](https://aiagentmemory.org/articles/ai-memory-injection/): AI memory injection plants malicious content in an agent's long-term memory so it misleads later sessions. Real attacks (MINJA, AgentPoison, SpAIware, MemForm, MemHarmBench) and defenses.
- [AI Memory MCP Servers: Official Server and Alternatives](https://aiagentmemory.org/articles/ai-memory-mcp-server/): What an AI memory MCP server is, how the official MCP memory server's knowledge graph works, and how Graphiti, Basic Memory, Mem0, MemPalace and Hindsight compare.
- [Azure AI Foundry Memory: Agent Service Long-Term Memory](https://aiagentmemory.org/articles/azure-ai-foundry-long-term-memory/): How long-term memory works in Azure AI Foundry Agent Service (Microsoft Foundry): memory stores, memory types, scope, TTL, quotas, regions, pricing and Python code.
- [Best AI Memory Framework: How to Choose by Use Case](https://aiagentmemory.org/articles/best-ai-memory-framework/): Which AI memory framework should you pick? A decision guide by use case: Mem0, Zep, Letta, Cognee, LangMem, Supermemory, Hindsight and Honcho, with prices.
- [Best Chatbot for Memory: Which AI Apps Remember You](https://aiagentmemory.org/articles/best-chatbot-for-memory/): Which chatbot has the best memory in 2026? A sourced comparison of ChatGPT, Claude, Gemini, Copilot and Character.AI: what each remembers, free plans, controls.
- [Best In-Memory Vector Databases: FAISS, Chroma, More](https://aiagentmemory.org/articles/best-in-memory-vector-database/): The best in-memory vector databases by language: FAISS, hnswlib, USearch, Chroma, Qdrant local mode, LangChain and Spring AI stores, with Python code.
- [Build a Chatbot with Memory in LangGraph (Python)](https://aiagentmemory.org/articles/chatbot-with-memory-langgraph/): How to build a chatbot with memory in LangGraph: a checkpointer for chat history, a store for long-term user memory, Postgres for production, and example repos.
- [Character.AI Memory Limit: Pins, Chat Memories, Lorebooks](https://aiagentmemory.org/articles/what-is-character-ai-memory-limit/): What is Character.AI's memory limit? Why characters forget, the real limits on pins (15 or 30) and Chat Memories (400 characters), and how to improve memory.
- [Chatbot Memory on GitHub: Open-Source Repos (2026)](https://aiagentmemory.org/articles/chatbot-memory-github/): A checked list of GitHub repos for chatbot memory: chat apps with built-in memory, memory libraries and templates, with license, stars, latest release and how each one remembers.
- [Claude AI Long-Term Memory: How It Works and Settings](https://aiagentmemory.org/articles/claude-ai-long-term-memory/): Does Claude have long-term memory? How Claude's memory and chat search work, plans, how to turn memory on, import from ChatGPT, context window sizes, and the API.
- [Claude Code Memory: CLAUDE.md, Auto Memory and Plugins](https://aiagentmemory.org/articles/claude-code-memory/): How Claude Code memory works: CLAUDE.md locations and load order, @imports, auto memory and how to turn it off, /memory, plus memory MCP servers, claude-mem and Hindsight.
- [claude-mem: How It Works, Install, Cost and Limits](https://aiagentmemory.org/articles/claude-mem/): What claude-mem is, how its hooks and local worker store and search your coding sessions, how to install it on Claude Code and other agents, what it costs, and when to pick something else.
- [Cline Memory Bank: Files, Setup, Cost and Limits](https://aiagentmemory.org/articles/cline-memory-bank/): What the Cline Memory Bank is, the six markdown files and their hierarchy, how to set it up with .clinerules, what it costs in context tokens, and when to use a real memory server.
- [CodeGraph: Code Knowledge Graph for Claude Code and Codex](https://aiagentmemory.org/articles/codegraph-memory/): What CodeGraph (colbymchenry/codegraph) does for Claude Code, Codex and Cursor: a local code graph over MCP, its benchmark numbers, limits, and when to add Hindsight.
- [Codex Memory: AGENTS.md, Memories and Add-ons](https://aiagentmemory.org/articles/codex-memory/): How OpenAI Codex memory works (0.166.0): AGENTS.md load order and 32 KiB limit, Codex memories (v2 by default for new users), default limits, config keys, importing Claude Code memory, and adding Hindsight.
- [Cognee Alternatives: 7 Agent Memory Tools Compared](https://aiagentmemory.org/articles/cognee-alternatives/): Cognee alternatives compared: Mem0, Graphiti, Zep, Letta, Hindsight, MemPalace and Supermemory. What each does, licenses, self-hosting and pricing, checked October 2026.
- [Context Window Limitations in LLMs: Research and Fixes (2026)](https://aiagentmemory.org/articles/context-window-limitations-solutions/): Why LLMs get worse as the context window fills (lost in the middle, RULER, NoLiMa, context rot), what long prompts cost, and when retrieval or memory beats full context, with October 2026 results.
- [CrewAI Long-Term Memory: The Unified Memory Class](https://aiagentmemory.org/articles/crewai-long-term-memory/): How CrewAI long-term memory works since v1.10: one Memory class with scopes, LanceDB storage, composite recall scoring, embedder setup, and how to reset it.
- [DeepSeek Harness Memory: Sessions, Compaction, Plugins](https://aiagentmemory.org/articles/deepseek-harness-memory/): How memory works in DeepSeek Harness (dsh): session logs, auto compaction, MCP memory overlays, and long-term memory plugins: Hindsight, Honcho, Mem0, ReMe, dsh-mnemon.
- [Dify LLM Memory: Node Memory, Window Size, Variables](https://aiagentmemory.org/articles/dify-llm-memory/): How memory works in Dify: LLM node memory and window size in Chatflow, conversation variables, Agent App history compaction, and plugins for long-term memory.
- [Embedding Models for RAG and Agent Memory (2026)](https://aiagentmemory.org/articles/embedding-models-for-rag/): How to choose an embedding model for RAG and agent memory: MTEB scores, dimensions, context length and cost for OpenAI, Gemini, Voyage, Cohere, Qwen3, BGE-M3.
- [Episodic Memory in AI Agents: How Agents Recall Events](https://aiagentmemory.org/articles/episodic-memory-in-ai-agents/): What episodic memory is in AI agents and LLMs, what an episode record holds, how Generative Agents and Reflexion use it, and how to build one with FAISS.
- [Everything Claude Code (ECC) Memory and Instincts](https://aiagentmemory.org/articles/everything-claude-code-memory/): How memory works in ECC (Everything Claude Code, affaan-m/ECC): session hooks, instincts with confidence scores, the Memory Vault, defaults, limits, and where Hindsight fits.
- [Extending the LLM Context Window: How It Works](https://aiagentmemory.org/articles/extending-llm-context-window/): How to extend an LLM context window: position interpolation, NTK and YaRN RoPE scaling, LongRoPE, long-context fine-tuning, ring attention and KV cache tricks.
- [Gemini AI Long-Term Memory: Past Chats and Settings](https://aiagentmemory.org/articles/gemini-ai-long-term-memory/): How Gemini's long-term memory works in 2026: memory from past chats, asking Gemini to remember, importing memory from ChatGPT or Claude, how to delete it, and context windows.
- [GitHub Copilot Memory: What It Stores and Where](https://aiagentmemory.org/articles/github-copilot-memory/): How GitHub Copilot Memory works: repository facts, user preferences, citation checking, the 28-day expiry, which Copilot surfaces use it, and how to manage or replace it.
- [Google Titans: How AI Learns to Memorize at Test Time](https://aiagentmemory.org/articles/google-titans-give-ai-human-like-memory/): Google Titans explained from the paper: a neural long-term memory that learns at test time using surprise, momentum and forgetting, its MAC/MAG/MAL variants and results.
- [Graphify: Codebase Knowledge Graph Memory for Agents](https://aiagentmemory.org/articles/graphify-memory/): What Graphify does for Claude Code, Codex and Cursor: a local tree-sitter knowledge graph of your repo, its memory benchmarks, limits, and when to add Hindsight.
- [Grok Bot Memory: How xAI's Bots Remember, and Plugins](https://aiagentmemory.org/articles/grok-bot-memory/): How Grok Bot memory works: what each Bot remembers, what all Bots share, what xAI hasn't documented, and the Hindsight, Supermemory and claude-mem plugins.
- [Headroom: Context Compression for Claude Code and Codex](https://aiagentmemory.org/articles/headroom-context-compression/): What Headroom (headroomlabs-ai/headroom) does for Claude Code, Codex and other agents: local context compression, its token numbers, its memory features and limits.
- [Hermes Agent Desktop: Install, Memory and Setup](https://aiagentmemory.org/articles/hermes-agent-desktop/): Hermes Agent Desktop is Nous Research's native app for Hermes. How to install it, how it shares memory and sessions with the CLI, the Memory Graph, and limits.
- [Hermes Agent Memory: Limits, Best Provider and Fixes](https://aiagentmemory.org/articles/hermes-agent-memory/): Hermes Agent remembers with MEMORY.md (2,200 chars) and USER.md (1,375 chars), loaded once per session, plus SQLite session search. Which of its 9 memory providers to pick, by need, and why Hermes forgets.
- [Hermes Agent Plugins: Install, Catalog, Memory Providers](https://aiagentmemory.org/articles/hermes-agent-plugins/): How Hermes Agent plugins work: plugin types, hermes plugins install, the SHA-pinned catalog, memory provider plugins, and what v0.23.0 changed (Nov 2026).
- [Hermes Agent Skills: How Procedural Memory Works](https://aiagentmemory.org/articles/hermes-agent-skills/): How Hermes Agent skills work: how the agent writes SKILL.md files from experience, where they live, how they load, the Skills Hub, the curator, and limits.
- [Hierarchical Memory for LLM Agents: Tiers and Trees](https://aiagentmemory.org/articles/llm-hierarchical-memory/): How hierarchical memory works in LLM agents: MemGPT-style tiers, MemoryOS, and abstraction trees like RAPTOR, H-MEM and Zep's graph layers, with code.
- [Hindsight AI Memory: How It Works and When to Use It](https://aiagentmemory.org/articles/ai-memory-hindsight/): Hindsight is an MIT-licensed agent memory server with retain, recall and reflect. How it works, its BEAM benchmark lead, its limits, and when to pick another tool.
- [Hindsight Memlake vs Postgres, turbopuffer and LanceDB](https://aiagentmemory.org/articles/hindsight-memlake/): Hindsight Memlake vs pgvector, turbopuffer, LanceDB and Pinecone for agent memory: architecture, writes, cold start, licensing, Vectorize-run numbers, limits.
- [Holographic Memory for AI Agents: HRR Explained](https://aiagentmemory.org/articles/holographic-memory-ai-agent/): What holographic memory means for AI agents: holographic reduced representations (HRR), how bind and unbind work, and the Hermes Agent holographic provider.
- [Honcho Memory (v3.5.0): How It Works, Self-Hosting, Pricing](https://aiagentmemory.org/articles/honcho-llm-memory/): Honcho by Plastic Labs: AGPL memory that models each user as a peer. $2 per 1M tokens, 90.4% LongMemEval (self-reported), self-hosting, and 3.5.0 source citations.
- [How AI Memory Works: Write, Store, Retrieve, Inject](https://aiagentmemory.org/articles/how-ai-memory-works/): How does memory work in AI? The model forgets, so the app saves notes, searches them and adds them to the prompt. ChatGPT, Claude, Gemini, Mem0, Letta and Hindsight compared step by step.
- [How LLM Memory Works: Weights, Context and KV Cache](https://aiagentmemory.org/articles/how-llm-memory-works/): How LLM memory works inside the model: knowledge in weights, the context window, the KV cache, why each API call is stateless, and where external memory fits.
- [How to Add Memory to a Chatbot (with Python)](https://aiagentmemory.org/articles/how-to-add-memory-to-chatbot/): How to add memory to a chatbot: chat history buffers, trimming, rolling summaries and long-term facts, with tested Python code for SQLite and LangChain.
- [How to Find ChatGPT Memory: View, Turn Off, Delete](https://aiagentmemory.org/articles/how-to-find-chatgpt-memory/): How to find ChatGPT memory in Settings, see what it remembers, turn memory on or off, and fully delete saved memories and chat history, on web and phone.
- [How to Give a Local LLM Memory (Ollama, SQLite, Python)](https://aiagentmemory.org/articles/how-to-give-local-llm-memory/): Local LLMs don't remember between chats. How to add persistent memory offline: chat apps, MCP, and tested Python with Ollama, SQLite, Mem0 and the Agno agent framework.
- [How to Give AI Agents Memory (Python Examples)](https://aiagentmemory.org/articles/how-to-give-ai-agents-memory/): How to give AI agents memory that lasts across sessions: the two main patterns, a 7-step build, and tested Python code with Pydantic AI, LangMem and Mem0.
- [Infinite Context Window LLM: Does It Exist?](https://aiagentmemory.org/articles/infinite-context-window-llm/): No LLM has an infinite context window. What Infini-attention, StreamingLLM, RMT and MemGPT actually do, and how chatbots fake infinite memory with storage.
- [Janitor AI Memory: Chat Memory, JLLM Context and Fixes](https://aiagentmemory.org/articles/janitor-llm-memory/): How Janitor AI memory works: JLLM's ~9K token context, permanent vs temporary tokens, the Chat Memory box, the August 2026 memory changes, and how to make bots remember.
- [Knowledge Graphs for AI Agent Memory: How They Work](https://aiagentmemory.org/articles/ai-memory-knowledge-graph/): How knowledge graph memory works for AI agents: extraction, entity resolution, temporal edges, Graphiti, Neo4j, Cognee and GraphRAG, versus vector memory.
- [LangChain Memory in 2026: What Replaced Buffer Memory](https://aiagentmemory.org/articles/llm-memory-langchain/): How LLM memory works in LangChain 1.x: ConversationBufferMemory is deprecated, so use LangGraph checkpointers for chat history and stores for long-term memory.
- [Langflow Memory Chatbot: Message History and Sessions](https://aiagentmemory.org/articles/memory-chatbot-langflow/): Build a memory chatbot in Langflow: Agent built-in memory, Message History, session IDs, external chat memory, and memory bases (SQLite by default since 1.13) for long-term recall.
- [Largest Context Window Open-Source LLMs (Local, 2026)](https://aiagentmemory.org/articles/largest-context-window-llm-open-source/): Open-weight LLMs with the largest context windows as of October 2026, how to run long context in Ollama and llama.cpp, and how much VRAM the KV cache needs.
- [Letta (formerly MemGPT): Memory, Setup and Pricing](https://aiagentmemory.org/articles/letta-ai-guide/): What Letta is today: the MemGPT successor, how its memory works (MemFS, memory blocks, dreaming, archival memory), how to use it from Python, and its pricing.
- [Letta Alternatives: 7 Agent Memory Options Compared](https://aiagentmemory.org/articles/letta-alternatives/): What Letta (formerly MemGPT) does now, why teams look for Letta alternatives, and how Mem0, Zep, Graphiti, Cognee, LangMem, Hindsight and Supermemory compare.
- [Letta vs LangChain Memory: Harness vs Library (2026)](https://aiagentmemory.org/articles/letta-vs-langchain-memory/): Letta vs LangChain memory in 2026: Letta Code's self-editing, git-backed memory files versus LangGraph checkpointers, stores and Deep Agents. When to pick each.
- [LLM Agent Memory Management: Write, Trim, Forget](https://aiagentmemory.org/articles/llm-agent-memory-management/): How to manage LLM agent memory: set a context budget, decide what to write, trim and summarize history, consolidate, expire with TTLs, delete, scope per user.
- [LLM Context Window Comparison: Largest Models (2026)](https://aiagentmemory.org/articles/context-window-llm-ranking/): Which LLM has the largest context window? A sourced comparison of context window sizes by model and provider as of October 2026, with long-context pricing.
- [LLM Context Window History: 512 Tokens to 1M+](https://aiagentmemory.org/articles/llm-context-window-history/): How LLM context windows grew from 512 tokens in GPT-1 to about 1M today: a dated timeline of models and the arxiv papers on attention and RoPE behind it.
- [LLM Context Window Optimization: A Practical Guide](https://aiagentmemory.org/articles/llm-context-window-optimization/): LLM context window optimization in practice: token budgets, compaction, tool result clearing, prompt caching, retrieval and memory, with real API settings.
- [LLM Memory Comparison: 11 Memory Systems Side by Side](https://aiagentmemory.org/articles/llm-memory-comparison/): LLM memory comparison table: Mem0, Zep/Graphiti, Letta, Cognee, LangMem, Supermemory, Hindsight, Honcho: license, storage, retrieval, hosting, benchmarks.
- [LLM Memory Compression: Compaction, Summaries, Tokens](https://aiagentmemory.org/articles/llm-memory-compression/): LLM memory compression explained: conversation compaction, fact extraction, prompt compression (LLMLingua), KV cache eviction and vector quantization.
- [LLM Memory Evaluation: Benchmarks for Agent Memory](https://aiagentmemory.org/articles/llm-memory-evaluation/): How to evaluate LLM and agent memory: LoCoMo, LongMemEval, MemoryAgentBench, BEAM, FEB-30, Bemula, SciValidityBench, BehaviorMemBench, MSC and DMR compared by size, tasks and metrics, plus how to run one.
- [LLM Memory Survey Papers: The 6 Worth Reading (2026)](https://aiagentmemory.org/articles/llm-memory-survey/): A guide to the main LLM and agent memory survey papers on arXiv, 2024 to 2026: what each covers, the taxonomy it proposes, and which one to read first.
- [LLM Memory With Obsidian: Karpathy's LLM Wiki Pattern](https://aiagentmemory.org/articles/llm-memory-obsidian/): How to use Obsidian as LLM memory, based on Andrej Karpathy's LLM Wiki gist: raw sources, an LLM-maintained wiki and a schema file, plus search and MCP tools.
- [LLM Parametric Memory: What Model Weights Remember](https://aiagentmemory.org/articles/llm-parametric-memory/): What parametric memory is in LLMs: how facts live in model weights, how much a model holds, how to update it (fine-tuning, ROME, MEMIT, Memory Tuning).
- [Mem0 Alternatives: 7 Agent Memory Tools Compared](https://aiagentmemory.org/articles/mem0-alternatives-compared/): Mem0 alternatives compared: Zep, Graphiti, Letta, Cognee, Hindsight, Supermemory and LangMem: licenses, self-hosting, memory models and pricing (Oct 2026).
- [Mem0 Paper: Production-Ready Agents with Long-Term Memory](https://aiagentmemory.org/articles/mem0-building-production-ready-ai-agents/): The Mem0 paper (arXiv 2504.19413) explained: extraction and update pipeline, the Mem0g graph variant, LoCoMo results, latency and tokens, and the Zep dispute.
- [Mem0 vs Cognee: User Memory vs Knowledge Graph](https://aiagentmemory.org/articles/mem0-vs-cognee/): Mem0 vs Cognee compared (Oct 2026): fact extraction vs knowledge graphs, open-source graph memory, local models, APIs, pricing and when to use each.
- [Mem0 vs Hindsight: Agent Memory Layers Compared](https://aiagentmemory.org/articles/mem0-vs-hindsight/): Mem0 vs Hindsight compared (Oct 2026): how each stores and retrieves memory, open-source vs hosted features, benchmarks and token cost, pricing, and which fits.
- [Mem0 vs Honcho: User Modeling, Self-Hosting and Benchmarks](https://aiagentmemory.org/articles/mem0-vs-honcho/): Mem0 vs Honcho (Oct 2026): facts vs peer reasoning, a self-hosting table (containers, LLM keys, auth defaults, telemetry), AGPL vs Apache 2.0, benchmarks, pricing.
- [Mem0 vs Letta: Memory Layer or Stateful Agent?](https://aiagentmemory.org/articles/mem0-vs-letta/): Mem0 vs Letta (MemGPT) compared in 2026: memory model, who decides what to store, Python support, self-hosting, pricing, and which one fits your agent.
- [Mem0 vs MemPalace: Extracted Facts vs Verbatim Recall](https://aiagentmemory.org/articles/mem0-vs-mempalace/): Mem0 vs MemPalace compared (Oct 2026): extracted facts vs verbatim local storage, LLM and API needs, coding agent hooks, benchmarks, and which one to choose.
- [Mem0 vs Supermemory: Agent Memory APIs Compared](https://aiagentmemory.org/articles/mem0-vs-supermemory/): Mem0 vs Supermemory compared (Oct 2026): memory model, RAG and connectors, open source vs hosted, self-hosting, benchmarks, pricing, and which one fits your app.
- [Memori by GibsonAI: SQL-Native Memory Storage for Agents](https://aiagentmemory.org/articles/ai-agents-memory-storage-memori-gibsonai/): Memori (GibsonAI, now Memori Labs) stores agent memory in your SQL database. How it works, Python setup, LoCoMo scores by category, and how it compares with Mem0 and Hindsight.
- [Memory Consolidation in AI Agents: Methods and Results](https://aiagentmemory.org/articles/memory-consolidation-ai-agents/): How LLM agents consolidate and forget memory: reflection, hierarchical summaries, fact merging, invalidation, sleep-time jobs, and October 2026 results (PreCoMem, REM).
- [MemoryOS: The Memory OS of AI Agent Paper Explained](https://aiagentmemory.org/articles/memory-os-ai-agent/): MemoryOS (EMNLP 2025) explained: three memory tiers, heat-based updates, LoCoMo results and the GitHub repo, plus how MemGPT/Letta and MemOS use the OS idea.
- [MemPalace: Milla Jovovich's Open-Source AI Memory](https://aiagentmemory.org/articles/mempalace-ai-memory-system/): MemPalace is an MIT-licensed local AI memory system co-created by Milla Jovovich and Ben Sigman. How wings, rooms and drawers work, and what its benchmarks measure.
- [Microsoft Agent Framework Memory: Sessions and Providers](https://aiagentmemory.org/articles/microsoft-agent-framework-memory/): How memory works in Microsoft Agent Framework (Python): AgentSession, history and context providers, and Mem0, Foundry, Cosmos DB, Redis, Hindsight compared.
- [Microsoft Memora: Agent Memory With Cue Anchors](https://aiagentmemory.org/articles/microsoft-memora-ai-memory-system/): Microsoft Memora is a research memory system for AI agents (ICML 2026). How primary abstractions and cue anchors work, its reported LoCoMo and LongMemEval results.
- [Mnemosyne vs Hindsight: Agent Memory Compared](https://aiagentmemory.org/articles/mnemosyne-vs-hindsight/): Mnemosyne vs Hindsight (Oct 2026): what Mnemosyne AI memory is, SQLite vs Postgres, recall, Hermes and MCP setup, BEAM scores from Mnemosyne's own README, and which fits.
- [n8n AI Agent Memory: Which Memory Node to Use](https://aiagentmemory.org/articles/best-memory-for-ai-agent-n8n/): The memory options for the n8n AI Agent node: Simple Memory, session IDs, Postgres, Redis, MongoDB and new Databricks Lakebase chat memory, deprecated Zep, long-term memory, clearing.
- [Open Source AI Memory Systems Compared: GitHub Repos](https://aiagentmemory.org/articles/open-source-memory-systems-compared/): Open-source AI memory systems compared: GitHub repos, licenses, stars (Oct 2026), languages and what each is good at. Mem0, Hindsight, MemPalace, Graphiti, Cognee.
- [OpenAI Agents SDK Memory: Sessions and Sandbox Memory](https://aiagentmemory.org/articles/openai-agents-sdk-memory/): How memory works in the OpenAI Agents SDK: Sessions for chat history (SQLite, Redis, SQLAlchemy, Conversations API), compaction, and sandbox agent memory files.
- [OpenClaw Memory: Files, Search and Embedding Models](https://aiagentmemory.org/articles/openclaw-ai-agent-memory/): How OpenClaw memory works: MEMORY.md and daily notes, hybrid memory_search, embedding models, dreaming, and 8 memory plugins compared (OpenViking, Hindsight).
- [OpenClaw vs Hermes Agent Memory: Side by Side](https://aiagentmemory.org/articles/openclaw-vs-hermes-agent-memory/): OpenClaw vs Hermes Agent memory compared: file layout, size limits, search, consolidation, external providers and migration, from each project's docs (Oct 2026).
- [OpenViking: ByteDance's Context Database for Agent Memory](https://aiagentmemory.org/articles/openviking-agent-memory/): OpenViking explained (Oct 2026): the viking:// filesystem, L0/L1/L2 layers, session memory extraction, Hermes and Claude Code plugins, benchmarks and how it compares to Hindsight.
- [Oracle AI Agent Memory: SDK, Setup, Limits](https://aiagentmemory.org/articles/oracle-ai-agent-memory/): Oracle AI Agent Memory (oracleagentmemory) explained: threads, context cards, long-term memory, search modes, TTL, setup in Python, and how it compares to Hindsight and Mem0.
- [Persistent Memory in AI: How It Works and How to Add It](https://aiagentmemory.org/articles/persistent-memory-ai/): What persistent memory in AI is, how apps and agents keep it across sessions and restarts, which tools offer it (AnythingLLM, ChatGPT, Mem0), and what breaks.
- [Procedural Memory in AI Agents: Prompts, Skills, Workflows](https://aiagentmemory.org/articles/ai-agent-procedural-memory/): Procedural memory in AI agents and LLMs: weights, code, prompts and skill files, how Voyager, AWM and Memp learn procedures, and LangMem prompt optimization.
- [RAG vs Agent Memory: Differences and When to Use Each](https://aiagentmemory.org/articles/rag-vs-agent-memory/): RAG vs agent memory: what retrieval-augmented generation is, how long-term memory differs (writes, updates, time, per-user scope) and how to combine both.
- [ReasoningBank: Agent Memory That Learns From Failures](https://aiagentmemory.org/articles/reasoningbank-llm-agent-memory-framework/): ReasoningBank (Google, ICLR 2026) explained: how agents distill reasoning strategies from wins and failures, MaTTS, WebArena and SWE-Bench results, and the code.
- [Semantic Memory in AI Agents: Facts, Profiles, Graphs](https://aiagentmemory.org/articles/semantic-memory-ai-agents/): What semantic memory is in AI agents: profiles vs collections, triples and graphs, extracting facts with LangMem, keeping facts current, and LongMemEval knowledge-update scores by system.
- [Short-Term Memory in AI Agents: Context, State, Limits](https://aiagentmemory.org/articles/short-term-memory-ai-agents/): What short-term memory is in AI agents, what it holds, why it runs out, and how to trim, summarize or offload it, with a LangChain checkpointer example.
- [Short-Term vs Long-Term Memory in Agentic AI](https://aiagentmemory.org/articles/short-term-and-long-term-memory-agentic-ai/): Short-term vs long-term memory in agentic AI: scope, storage, limits and failures side by side, how data moves between them, and a LangGraph code example.
- [Spring AI Chat Memory: ChatMemory, Advisors, JDBC](https://aiagentmemory.org/articles/spring-ai-conversational-memory/): How conversational memory works in Spring AI 2.0: ChatMemory, MessageWindowChatMemory, advisors, JDBC/Redis/Neo4j repositories, 2.0 breaking changes, and the 2.1 move to session memory.
- [Supermemory Alternatives: 7 Agent Memory Tools Compared](https://aiagentmemory.org/articles/supermemory-alternatives/): Supermemory alternatives compared (Oct 2026): Hindsight, Mem0, Zep, Graphiti, Letta, Cognee, Honcho. Licenses, self-hosting, memory models, benchmarks, pricing.
- [Supermemory: Long-Term AI Memory as a Service](https://aiagentmemory.org/articles/supermemory-long-term-ai-memory-as-a-service/): What Supermemory is: a hosted memory API with fact extraction, user profiles, hybrid RAG search and connectors. How it works, Python SDK, local mode, pricing.
- [Temporal Reasoning in AI Memory: Time-Aware Recall](https://aiagentmemory.org/articles/temporal-reasoning-ai-memory/): Why AI memory gets time wrong, the temporal questions agents face, bi-temporal graphs and time-aware retrieval, benchmark results, and a Graphiti example.
- [TencentDB Agent Memory (v2.0.3): Layers, Team Memory, Setup](https://aiagentmemory.org/articles/tencentdb-agent-memory/): TencentDB Agent Memory explained (v2.0.3, Oct 2026): what 2.0.3 changes, L0-L3 memory layers, Team Memory and Memory Hub, proxy setup for Claude Code and OpenClaw, and how it compares to Hindsight and Mem0.
- [Types of AI Agent Memory: Working, Episodic, Semantic](https://aiagentmemory.org/articles/ai-agents-memory-types/): The types of AI agent memory: working, episodic, semantic and procedural (CoALA), parametric vs external memory, and how LangGraph, Letta and Mastra name them.
- [Vector Databases for LLM Memory: How to Choose One](https://aiagentmemory.org/articles/vector-database-for-llm-memory/): How vector databases store long-term memory for LLM agents, what memory needs beyond RAG, and how pgvector, Qdrant, Chroma, Milvus, Pinecone and Redis compare.
- [Vertex AI Agent Engine Memory Bank: Guide and Pricing](https://aiagentmemory.org/articles/vertex-ai-agent-engine-memory-bank/): Vertex AI Agent Engine Memory Bank, now Agent Platform Memory Bank: how memory generation, scopes, topics, profiles and revisions work, ADK code, and 2026 pricing.
- [Voice Agent Memory: How Voice AI Remembers Callers](https://aiagentmemory.org/articles/voice-agent-memory/): How to give a voice agent long-term memory: the latency budget, recall per turn vs per call, and the options (VoiceMem, Mem0, Zep, Hindsight) for Pipecat, LiveKit and Vapi.
- [What Is Context Engineering? Techniques and Memory](https://aiagentmemory.org/articles/context-engineering/): Context engineering decides which tokens an AI agent sees at each step. The main techniques compared across Anthropic, LangChain, Manus and Breunig, and memory's role.
- [What Is Mem0? How the Mem0 Memory Layer Works](https://aiagentmemory.org/articles/what-is-mem0-ai/): What Mem0 is in 2026: how it extracts and stores memories, open source vs Platform, graph memory, setup, delete, and the new Free + bring-your-own-key pricing (Oct 26).
- [What Is Zep Memory? Zep Cloud, Graphiti and Pricing](https://aiagentmemory.org/articles/what-is-zep-memory/): What Zep memory is in 2026: the Zep Cloud temporal context graph, open-source Graphiti, how it stores and retrieves memory, outside test results, pricing, self-hosting, integrations.
- [Why Is AI Memory So Bad? Why Chatbots Forget](https://aiagentmemory.org/articles/why-is-ai-memory-so-bad/): Why is AI memory so bad? Chatbots forget because models don't learn from your chats, context windows run out, and memory features save only a selection.
- [Zep Alternatives: Open-Source and Hosted Options](https://aiagentmemory.org/articles/zep-alternatives/): Zep alternatives compared, including open-source options after Zep Community Edition: Graphiti, Mem0, Letta, Cognee, Hindsight and Supermemory (Oct 2026).
- [Zep Docker in 2026: Self-Hosting Options Explained](https://aiagentmemory.org/articles/zep-memory-docker/): Can you run Zep memory with Docker? Zep Community Edition is deprecated. The self-hosting options now: Graphiti server and MCP containers, legacy CE, BYOC.
- [Zep vs Mem0: Temporal Graph vs Fact Memory](https://aiagentmemory.org/articles/zep-memory-vs-mem0/): Zep vs Mem0 compared (Oct 2026): temporal context graph vs extracted facts, open source, self-hosting, APIs, pricing, the benchmark dispute, how to pick.