# Where Claude Cowork memory lives, and what open source keeps instead

> Yes, Claude Cowork has memory, but only when the session runs in the cloud: Anthropic's help center says sessions that run locally on your computer don't use memory. Kortix is the open-source CompanyOS, my pick, where project memory is files in a repository you own; I work at Kortix, so weigh that accordingly.

URL: https://opensourceclaudecowork.com/claude-cowork-memory

## Where memory lives, and who reads each folder

`Memory`

```
Memory folder Who reads it
Memory in My files Only you and your sessions.
Each agent keeps its notes about you in its own subfolder.
Memory in a channel’s folder The channel’s members.
Channels are experimental and off by default.
The project repository can hold memory too:
a memory/ folder with MEMORY.md as its index, for what the project is and how it works.
Agents edit it with their file tools, and every edit lands through a change request.
```

Source: https://kortix.com/docs/guide

The guide's memory section: which folder each person reads, and the `memory/` folder in the project repository that a change request reviews.

## What Anthropic's help center says about memory

Anthropic's [article on chat search and memory](https://support.claude.com/en/articles/11817273-use-claude-s-chat-search-and-memory-to-build-on-previous-context) says this, as of October 20, 2026: "Memory is shared between Chat and Claude Cowork when Cowork runs in the cloud; Cowork sessions that run locally on your computer don't use memory."

Anthropic states the boundary a second time: "Memory across Cowork and chat only works when Cowork runs in the cloud. It isn't available in Cowork sessions that run locally on your computer." The local case is where the feature stops. Summaries that quote only the first half give the wrong answer to anyone whose session runs on their own machine.

The article also says what memory is made of. Claude saves it as individual topics while you chat, rather than writing a summary after the fact, and each project carries its own. The setting is fixed when a task starts, so a task that begins without memory cannot pick it up later. On Team and Enterprise plans, "owners control whether memory is available, and it's off for each member until they turn it on." Memory is switched on by default for Free, Pro, and Max plans, which makes it the default experience for an individual and an opt-in one for a team.

## Where the session runs decides whether memory applies

Anthropic's [Get started with Claude Cowork](https://support.claude.com/en/articles/13345190-get-started-with-claude-cowork) explains the placement. "Cowork runs your sessions remotely in the cloud, so your sessions and files live with your Claude account and follow you across desktop, web, and mobile." Its mechanics section adds that the work "runs on Anthropic's servers, in an isolated environment, and your sessions and files are saved to your Claude account."

Cloud is the ordinary case today, so the cloud half of the memory article is the half that applies.

The product's edges are still moving. The Cowork product page carries the banner "Claude Cowork is now just Claude", and the help center describes Claude deciding whether a request gets a quick answer or a task ([the Claude Help Center](https://support.claude.com/en/articles/16761823-claude-cowork-and-chat-are-one-claude)). Memory behaves as before: it belongs to the account, and it is present when the session runs in the cloud.

## Three questions that separate account memory from company memory

A team's question is different: what does the company know after six months of work, and can anyone open that knowledge, review it, and hand it on?

**Who can read it.** Memory settings sit on each person's own Settings → Memory screen. On Team and Enterprise plans, a member begins with memory off until they turn it on. What a colleague learned last month is not a document anyone else can open.

**Where it lives.** With the Claude account, on Anthropic's servers, for a session that runs in the cloud. For a session on your own machine, the article says there is no memory to read at all.

**What happens when someone leaves.** A memory built out of one person's chats is held with that account. There is no file to pass to whoever picks up the work, so the context goes with the person.

Account memory does what it says, and its limit is the account: a colleague reads it only after turning it on, and the record does not outlast the person who built it.

## What owned memory looks like in Kortix

In Kortix, memory is a folder. The [Kortix docs](https://kortix.com/docs) describe a project as a git repo with `kortix.yaml` at its root, where agents, skills, memory, tools, triggers and connectors are files. The guide says the same in its memory section: the project repository can hold memory too, a `memory/` folder with `MEMORY.md` as its index, and agents edit it with their file tools. An edit is an ordinary file write, so what one session writes reaches the next person's session in the same repo.

Because memory is a file, a change to it arrives the way every other change does. "Session work reaches `main` through a change request. Merge is default-deny for agents." A person reads the diff before it lands, and that review is where a wrong lesson gets caught before it becomes policy. Who can read the folder is an ordinary permission question: "Per-resource permissions for people and agents. Roles, groups, and an audit trail." [The security and approvals page](/claude-cowork-security-and-approvals) takes the gate apart.

Placement stays a choice. "Run it on Kortix Cloud, in your VPC, or on your own on-prem network. Self-host is free." And the property that matters when a vendor renames a product, merges it, or changes its plan: "Open source: read it, fork it, audit it." What that covers is set out on [the page about whether Claude Cowork is open source](/is-claude-cowork-open-source).

Removing a person is then a permissions change, and their notes stay in the repo with every earlier version of each memory file. An account setting keeps no such record: it holds a switch position, not a history a reviewer can read.

## When account-level memory is enough

For one person working alone, Anthropic's arrangement is the shorter path. Cloud sessions only, and the switch already on. There is no colleague to hand context to and nothing that needs a reviewer, and the memory is genuinely personal: how you like a draft, who your manager is, which deadline moved. Building a repository to hold that would be work with no reader.

The calculation changes with the second person. Two members, each starting with memory off until they turn it on, make a company whose shared knowledge depends on settings nobody audits.

## Four questions the memory answer raises

### Does Claude AI have a memory feature?

Yes. Claude builds memory from your chats and saves it as individual topics, each one added as you go. On Team and Enterprise plans an owner decides whether the feature is available.

### Is Claude Cowork different from regular Claude?

Less than it used to be. The Cowork product page says "Claude Cowork is now just Claude", and Claude now decides whether a request gets a quick answer or a task. Cloud sessions, scheduled tasks and shared memory describe the same product.

### Does Claude Cowork access your files?

Yes. On desktop it reads and writes local files without uploads, and in a cloud session it works on the files you hand it and saves the output to your account. Anthropic's Cowork article documents both.

### Does Claude Cowork run in the background?

Yes, in the cloud. Anthropic says a cloud session keeps working when you close your laptop, and scheduled tasks run with no device online. A task that needs something on your machine reaches it through the desktop app.

## Put the memory where your company can keep it

Start on Kortix Cloud, or self-host the whole stack on your own box.

[Try Kortix](https://kortix.com) · [Read the data privacy page](https://opensourceclaudecowork.com/claude-cowork-data-privacy)

---

The self-hosting handbook: a guide to open-source alternatives to Claude Cowork, built around three checks, what you can read, what you can host, and what you can review. Kortix, the open-source CompanyOS, is my pick, and I work at Kortix, so weigh that accordingly.

Revised 2026-10-25

I write this site and work at Kortix, so weigh that accordingly. Claude, Claude Cowork, Claude Code, ChatGPT, Gemini, Copilot and the product names on this site belong to their owners, and this site names them only to compare.

© 2026 opensourceclaudecowork.com
