Is there a memory plugin for Claude, and which kind do you need?

Short answer

Yes, in three forms. A custom connector links Claude to a remote MCP memory server and works in Claude, Cowork and Claude Desktop on every plan (Free: one connector). A desktop extension runs a local MCP memory server on your own computer. A plugin bundles skills, connectors and sub-agents into one install. None of them changes Claude's built-in memory; they give Claude tools to read and write memory elsewhere.

Three ways to add memory to Claude

“Memory plugin” is used for three different things in Claude, and they install in different places.

Custom connector Desktop extension Plugin
What it is A remote MCP server Claude reaches over the internet A local MCP server packaged as a .mcpb file A bundle of skills, connectors and sub-agents
Where it runs On a server you or a vendor host On your own computer Its parts run where each part runs
Works in Claude (web), Cowork, Claude Desktop Claude Desktop Chat, Cowork; synced to Claude Code
Plans Free (one connector), Pro, Max, Ultra, Team, Enterprise Desktop app users; admin controls on Team and Enterprise All paid plans
Install Customize > Connectors > “+ Add” > “Add custom connector” Settings > Desktop app > Extensions > Manage Customize > Plugins > Discover

MCP, the Model Context Protocol, is the open standard that lets Claude call tools on another system. A memory server is simply an MCP server whose tools are “save this” and “find what we know about that”.

Option 1: a custom connector (remote memory)

This is the route when memory must be reachable from the browser, from several devices, or by colleagues. On Pro and Max you add it yourself under Customize > Connectors. On Team only Owners and Primary Owners can add it to the organization (on Enterprise also people with a custom role that includes Manage access to Libraries); members then connect under Customize > Connectors. Two technical requirements trip people up: Claude connects from Anthropic’s cloud, not from your computer, so the server must be reachable on the public internet on port 443; servers behind a VPN or firewall need Anthropic’s IP addresses allowlisted.

Option 2: a desktop extension (local memory)

A desktop extension is a single-click package for a local MCP server, so you do not have to edit JSON configuration by hand. Node.js, Python and binary servers are supported, and Claude Desktop ships its own Node.js. The MCP server runs locally; where the memory is stored and who can reach it depend on that server’s implementation. A local extension does not work in claude.ai in the browser. For the most common local option, see setting up the official server-memory MCP in Claude Desktop.

Option 3: a plugin

A plugin bundles skills, connectors and sub-agents for a role or team, and is saved to your account, so it also appears in Claude Code when you sign in there. A memory plugin typically contains a connector to the memory server plus a skill that tells Claude when to save and when to look things up. On Team and Enterprise an owner can enable sharing under Organization settings > Plugins & skills > Policy, and members can publish a plugin to the organization. For the Claude Code side, see how Claude Code plugins can add shared memory.

The misunderstanding: a plugin upgrades Claude’s own memory

None of these options changes the memory you see under Settings > Your Claude. That built-in memory stays personal, stored as topics, split per Project. A connector or plugin gives Claude extra tools; Claude decides when to call them, and only in chats where the connector is switched on (via “+” > Connectors). If you want the external memory used every time, say so in project instructions or in a skill.

Check safety before you connect. Anthropic does not verify custom connectors. It advises connecting only to servers you trust, disabling tools you do not need, choosing “Allow always” only for trusted tools, and watching for prompt injection in tool results.

What this means for a team

Built-in memory cannot be shared between colleagues, so a team that wants shared memory in Claude has to go through a connector or plugin. The choice matters: a local extension gives each person their own file again, while a remote connector added by the owner gives everyone the same server. Then the real questions start: who may see what, how corrections from one person reach the others, and whether ChatGPT, Copilot and your coding agents can use the same memory. Background: best MCP memory server for a team.

Where Bombos fits

Bombos is that kind of plugin, built for a team rather than one person: it gives every AI tool your team uses, Claude, Cowork and Claude Code as well as ChatGPT, Copilot, Grok, Cursor and your own agents, one shared memory with team knowledge, rules learned from corrections and what already lives in mail, documents and repos, with every answer traceable to its source. It runs in the cloud or on your own server, and one person can try it first for their own agents.

Related questions

Does Claude already have memory without a plugin?

Yes. Claude has built-in memory on every plan, stored as topics, with a separate memory per Project. A plugin or connector adds memory outside Claude, for example one that other tools can also use.

Can a memory plugin read my Claude memory?

Anthropic does not document an API for that. You can view and copy your memory under Settings > Your Claude, or ask Claude to write it out verbatim.

Can I use a local memory server in claude.ai in the browser?

No. Desktop extensions run local MCP servers in Claude Desktop, and plugins with local MCP servers run in Cowork and Claude Code, not in chat. For the web you need a remote server added as a custom connector.

Is a third-party memory server safe?

Anthropic does not verify custom connectors. Connect only to servers you trust, review the permissions, and watch for prompt injection in what the tools return.

Sources

  1. Get started with custom connectors using remote MCP (Claude Help Center)
  2. Getting started with local MCP servers on Claude Desktop (Claude Help Center)
  3. Use plugins in Claude (Claude Help Center)
  4. Use Claude's chat search and memory (Claude Help Center)
The first half hour · free

What does your AI forget?

  1. 01You tell us which AI you use and where it keeps starting over.
  2. 02We show what a shared memory changes there, on your own examples.
  3. 03You keep the plan, with or without Bombos.
Piet Baudoin
Piet BaudoinYou talk to Piet, from the Bombos teamDutch team from Utrecht and Nijmegen · data in Europe
We reply the same working day.