# Memlo

> Memlo is a memory service for AI agents. Agents save short notes, called memories, and find them again later with search. One person's memories work across all of their agents and apps.

Memlo is reached over MCP at https://memlo.ai/mcp (Streamable HTTP, OAuth sign-in). An AI assistant asked to set Memlo up for someone should read https://memlo.ai/agent.md first.

Agents using Memlo used 22% fewer tokens than with file-based memory, on average, in Memlo's own measurement from a year of real use. How it was measured: https://memlo.ai/guides/reduce-agent-token-usage#how-we-measured

## Docs

- [Set up Memlo for someone](https://memlo.ai/agent.md): setup steps per client, sign-in, the tools, good habits, limits and troubleshooting
- [What Memlo is](https://memlo.ai/agent#what-memlo-is)
- [Before you start: an account](https://memlo.ai/agent#before-you-start-an-account)
- [Setup at a glance](https://memlo.ai/agent#setup-at-a-glance)
- [Set up Claude Code](https://memlo.ai/agent#set-up-claude-code)
- [Set up claude.ai or Claude Desktop](https://memlo.ai/agent#set-up-claude-ai-or-claude-desktop)
- [Set up ChatGPT](https://memlo.ai/agent#set-up-chatgpt)
- [Set up the Gemini web app](https://memlo.ai/agent#set-up-the-gemini-web-app)
- [Set up another MCP client](https://memlo.ai/agent#set-up-another-mcp-client)
- [What sign-in looks like](https://memlo.ai/agent#what-sign-in-looks-like)
- [Check that it works](https://memlo.ai/agent#check-that-it-works)
- [Load Memlo at the start of every session](https://memlo.ai/agent#load-memlo-at-the-start-of-every-session)
- [The tools](https://memlo.ai/agent#the-tools)
- [Good habits](https://memlo.ai/agent#good-habits)
- [Limits and spend](https://memlo.ai/agent#limits-and-spend)
- [Early access](https://memlo.ai/agent#early-access)
- [Troubleshooting](https://memlo.ai/agent#troubleshooting)

## Guides

- [All guides](https://memlo.ai/guides): the index
- [What is an MCP server, and how do you add one to Claude, ChatGPT and Codex?](https://memlo.ai/guides/what-is-an-mcp-server): What an MCP server is, how the Model Context Protocol works (tools, resources, transports, sign-in), and how to add a server to Claude, ChatGPT, Codex and VS Code.
- [How to give Claude Code memory that lasts between sessions](https://memlo.ai/guides/claude-code-memory): How to give Claude Code a memory that lasts between sessions with the Memlo MCP server: the command, sign-in, a SessionStart hook and the tools.
- [How to add a memory MCP server to Codex](https://memlo.ai/guides/codex-memory): How to add an MCP server to OpenAI Codex CLI with codex mcp add, using a memory server as the example: one command, OAuth sign-in, and a line in AGENTS.md.
- [How to connect a remote MCP memory server to any AI app](https://memlo.ai/guides/mcp-memory-server): How to connect a remote MCP memory server to claude.ai, Claude Desktop, ChatGPT, Gemini, VS Code or any MCP client, using Memlo as the example.
- [Why your coding agent forgets between sessions, and how to fix it](https://memlo.ai/guides/why-coding-agents-forget): Why Claude Code, Codex and other coding agents forget what you told them in the last session, and four ways to carry context across sessions.
- [CLAUDE.md and AGENTS.md vs a memory server: which one you need](https://memlo.ai/guides/claude-md-vs-memory-server): A fair comparison of CLAUDE.md, AGENTS.md and an MCP memory server for coding agents: what each does well, their limits, and when files are enough.
- [How to share one memory between Claude Code, Codex and other AI agents](https://memlo.ai/guides/share-memory-between-agents): How to give Claude Code, Codex, ChatGPT and other AI agents one shared memory over MCP, keep it organized in spaces, and share it with a team.
- [How to cut the tokens your coding agent spends finding context](https://memlo.ai/guides/reduce-agent-token-usage): Ways to cut the tokens Claude Code, Codex and other coding agents spend on context: shorter instruction files, scoped rules, clear sessions and memory search.
- [What CLAUDE.md is and how Claude Code uses it](https://memlo.ai/guides/claude-md): What CLAUDE.md is, where Claude Code looks for it, how it loads, how long it should be, and how it differs from auto memory and AGENTS.md.
- [What AGENTS.md is and how coding agents use it](https://memlo.ai/guides/agents-md): What AGENTS.md is, which coding agents read it, where Codex looks for it, how nested files combine, its size limit, and how to share it with Claude Code.
- [How to add an MCP server to Claude Code](https://memlo.ai/guides/claude-code-mcp-servers): Add MCP servers to Claude Code with claude mcp add: remote HTTP and local stdio servers, scopes, .mcp.json, sign-in with /mcp, and fixes for common errors.
- [claude-mem vs Memlo: which memory to use with Claude Code](https://memlo.ai/guides/claude-mem-vs-memlo): A fair comparison of claude-mem and Memlo for Claude Code memory: how each saves memories, where data lives, which agents connect, price and when to pick each.
- [Mem0 vs Memlo: which memory to use](https://memlo.ai/guides/mem0-vs-memlo): A fair comparison of Mem0 and Memlo: who each is built for, SDK and MCP access, self-hosting, how memories are made, pricing, and when to pick each one.
- [Supermemory vs Memlo: which memory to use](https://memlo.ai/guides/supermemory-vs-memlo): A fair comparison of Supermemory and Memlo: MCP setup, SDKs, connectors, self-hosting, how memories are made, teams, pricing and when to pick each one.
- [How to add a memory MCP server to ChatGPT](https://memlo.ai/guides/chatgpt-mcp-server): How to add a custom MCP server to ChatGPT, using a memory server as the example: the plan you need, the steps on chatgpt.com, OAuth sign-in and using it in a chat.
- [Keep CLAUDE.md short: move the rest into shared memory](https://memlo.ai/guides/keep-claude-md-short): How to shrink a long CLAUDE.md or AGENTS.md: keep the rules every task needs, and move decisions and notes into a shared memory your agents search.
- [Shared memory for Claude Code subagents and agent teams](https://memlo.ai/guides/claude-code-subagents-shared-memory): Claude Code subagents and agent teams each start with their own context. How to give them one shared memory, a handoff log, and edits that never silently overwrite.
- [Claude Code Projects shared memory vs Memlo](https://memlo.ai/guides/claude-code-projects-vs-memlo): Claude Code Projects gives parallel cloud threads a shared memory. How it compares with Memlo across vendors, local sessions and people, and which to use.
- [Context engineering for coding agents: search memory instead of loading CLAUDE.md](https://memlo.ai/guides/context-engineering-coding-agents): Context engineering for Claude Code and Codex: load less at session start, let the agent search a memory for what each step needs, and keep notes editable.
- [What is AI agent memory? How long-term memory for LLM agents works](https://memlo.ai/guides/ai-agent-memory): What long-term memory for AI agents is, the ways it's built (files, vector search, knowledge graphs, memory servers), how agents write to it, and what to watch for.
- [How to add a custom connector to Claude](https://memlo.ai/guides/claude-custom-connector): How to add a custom connector to claude.ai and Claude Desktop: the steps, plan limits, OAuth sign-in, the URL mistake to avoid, and a memory server as the example.
- [Graphiti vs Memlo: which to use for agent memory](https://memlo.ai/guides/graphiti-vs-memlo): A fair comparison of Graphiti and Memlo: an open-source temporal knowledge graph framework you host, or a hosted memory server your agents share. When to pick each.
- [Zep vs Memlo: which memory to use](https://memlo.ai/guides/zep-vs-memlo): A fair comparison of Zep and Memlo: Zep's context graph platform and SDKs for building agents, or a hosted memory server your own agents share.
- [Cognee vs Memlo: which memory to use](https://memlo.ai/guides/cognee-vs-memlo): A fair comparison of Cognee and Memlo: an open-source graph and vector memory engine you can self-host, or a hosted memory server your agents share.
- [Letta vs Memlo: which to use for agent memory](https://memlo.ai/guides/letta-vs-memlo): A fair comparison of Letta (formerly MemGPT) and Memlo: Letta's stateful agents with git-backed memory, or a hosted memory server your existing agents share.
- [Basic Memory vs Memlo: which memory to use](https://memlo.ai/guides/basic-memory-vs-memlo): A fair comparison of Basic Memory and Memlo: local Markdown notes with an MCP server and optional cloud, or a hosted memory server your agents share.
- [What is a memory bank for coding agents?](https://memlo.ai/guides/memory-bank): What a coding-agent memory bank is: the Markdown files and instructions behind Cline's Memory Bank and Cursor setups, their limits, and the alternatives.
- [How memory works in Cursor today](https://memlo.ai/guides/cursor-memory): How memory works in Cursor today: Memories were removed in version 2.1, rules and AGENTS.md took their place, plus memory-bank setups and MCP memory servers.
- [What are Cursor rules, and how do you use them?](https://memlo.ai/guides/cursor-rules): How Cursor rules work: project rules as .mdc files in .cursor/rules, the four ways a rule applies, User Rules, Team Rules, AGENTS.md, and what to keep out of rules.
- [How ChatGPT memory works, and how to share it with other agents](https://memlo.ai/guides/chatgpt-memory): How ChatGPT memory works: saved memories, chat history, project memory, Temporary Chat and controls, plus how to share what ChatGPT knows with Claude and Codex.
- [Claude memory vs ChatGPT memory: how they compare](https://memlo.ai/guides/claude-memory-vs-chatgpt-memory): A fair, dated comparison of Claude's and ChatGPT's memory: what each remembers, plans, project memory, private chats, export, and where shared memory fits.
- [What are Claude Projects, and how does their memory work?](https://memlo.ai/guides/claude-projects): What Claude Projects are: knowledge, instructions, per-project memory, RAG for large projects, sharing on Team plans, and where a shared memory fits.
- [What is Serena MCP, and how does it work with a memory server?](https://memlo.ai/guides/serena-mcp): What Serena MCP is: symbol-level code tools through language servers, adding it to Claude Code, its project memories, and how it pairs with a memory server.
- [LangMem vs Memlo: which memory to use](https://memlo.ai/guides/langmem-vs-memlo): A fair comparison of LangMem and Memlo: LangChain's open-source memory library for LangGraph agents, or a hosted memory server your existing agents share.

## Pages

- [Home page and waitlist](https://memlo.ai/)
- [How it works](https://memlo.ai/how-it-works)
- [Guides](https://memlo.ai/guides)
- [The app](https://memlo.ai/app)
- [Terms](https://memlo.ai/legal/terms)
- [Privacy](https://memlo.ai/legal/privacy)
- [Acceptable use](https://memlo.ai/legal/acceptable-use)

## Optional

- [Full text](https://memlo.ai/llms-full.txt): the setup page and every guide in one file