Alcoven

Can a client ask my Claude project about the work themselves?

Blog · Published 28 October 2026

Only if they join your Claude organisation. Claude shares a project only with members of the same Team or Enterprise organisation, so a client would need a seat in yours. Without one, you can share a snapshot of a chat or an artifact, as your plan allows: they can read the work, but they can’t ask your Claude about it.

So the client ends up going through you to reach your AI. One client, writing to the person whose Claude had done the work, put it plainly (translated from Turkish): “If I had your Claude, I’d tell it myself, and I’d check the outputs too.” And: “The one thing I’m asking of you is to check the outputs.”

Here is what Claude allows today, as Anthropic’s help pages describe it, and what you can do instead.

Can you share a Claude project with someone outside your organisation?

No. Project sharing is on Claude’s Team and Enterprise plans, and the Share menu adds “people from your organization by name or email address”. Enterprise can share a project with a group, and even then “a shared project stays scoped to its own organization.” On Free, Pro and Max, a project is yours alone.

The new projects in Claude Code, in beta on Pro and Max, don’t change this. Anthropic’s docs say: “A project belongs to one user. You can’t share a project or its threads with another user.”

What would a client need to use your Claude project?

A seat in your organisation. On a Team or Enterprise plan:

  1. An admin adds the client under Organization settings > Members. Their email address “must use one of your organization’s allowed email domains”. An owner can add the client’s domain under Organization settings > Identity and access, unless your organisation uses domain capture.
  2. The client takes a seat, which your organisation pays for. A pending invitation holds a seat as soon as it is sent, and a Team seat “is assigned to one named individual, and only that person may use it”, so you can’t hand the client your own login.
  3. You share the project with them with Can view access. They “can see project contents, knowledge, and instructions, and chat within the project, but cannot edit it.”

Even then, your own chats stay yours. Anthropic is plain about it: “your chats within that project will be private and inaccessible to other members of your organization unless you manually share them.” So the client’s chats start from the project’s files and instructions. What you worked out in your own chats reaches them only if you share a chat or add it to the project’s knowledge.

Whether a client belongs inside your Claude organisation, under its settings and data rules, is for whoever runs it to decide.

What can you share with a client who has no seat?

Snapshots, artifacts and copies of your files. Each lets the client read the work. None lets them ask your Claude about it.

What you share Plans What the client can do
A chat, invited by email Free, Pro and Max; on Team and Enterprise only with people in your organisation Read a snapshot of the messages so far, with a free Claude account. They “can’t reply, copy it into their own account, or continue the conversation”
A chat, by public link Free, Pro and Max Read the snapshot, without the files you attached. No account needed
An artifact, invited by email (beta) Pro and Max; Team and Enterprise where an owner allows invitations outside the organisation (on by default on Team) View it, comment on it or edit it, as you choose, with a Claude account. “Parts of the artifact that use Claude or your connectors don’t work for them”

A shared chat is frozen at the moment you share it: messages after that stay private until you update the snapshot. When you invite someone outside your organisation by email, the files you attached aren’t included, and neither is the raw data from connectors or MCP tool calls. They see Claude’s final responses.

You can also send the client the files themselves. Projects are on every Claude plan, Free included, so they can add your files to a project of their own and ask their own Claude about them. Their Claude then knows the files as of the day you sent them, and none of your chats.

Can you ask a colleague’s Claude about their work yourself?

Only as far as they share it with you. Their chats are theirs, and so is their memory: on Team and Enterprise, once an owner turns memory on, “individual users manage their own memory settings”, and owners “can’t view or edit a user’s individual memories”. What you can reach depends on where the work lives.

  • In a shared Claude project. If you’re in the same Team or Enterprise organisation, ask them to share it. You get its knowledge and instructions and can chat with Claude in it. Their chats in it stay private unless they share one, and a shared chat is a snapshot you can read but can’t continue. More on this in sharing a Claude project with a colleague.
  • In a repository. If the work is code in git, clone it and ask your own Claude Code. Claude Code reads the repository’s CLAUDE.md, which Anthropic’s docs describe as “Team-shared instructions for the project”, shared through version control. You start from the same code and instructions; their conversations stay with them.
  • Only in their chats. Then every question goes through them, and the way out is a written handover your own AI can read: what it does now, how it was built, what was tried and what is still open.

What does checking the outputs take?

To check an AI’s output yourself, you need four things:

  1. The output as it stands now, with its earlier versions.
  2. The brief and the instructions it was made from.
  3. What the person asked their AI for, and what the AI read.
  4. A place to say what is wrong, where the next round of work will see it.

Claude’s sharing covers parts of this. A shared chat shows what was asked and what Claude answered, at one moment. A seat with Can view access adds the project’s files and instructions. For the fourth, Claude’s help centre says context “is not shared across chats within a project unless the information is added into the project knowledge base”, and adding to it takes Can edit access.

Where does Alcoven fit?

Alcoven keeps the project outside any one person’s AI: the brief, documents, tasks and history, shared by the people and agents on it. What your AI makes there goes on the project under your name. You invite a client by email to read it; readers join free on any plan, and you can close folders to them. In the Alcoven app, they ask its agent about the work; it answers from the project and cites what it read. Each document keeps every past version, and on a passage an AI wrote, Why this? shows who put it there, with which AI and when, and, when their AI passed it on, what they asked and what it read. When something is wrong, the client leaves a note that asks you, and it stays open until someone closes it with what came of it. They can also open the project in their own AI, where their company allows a connector. Their questions in the app count against your team’s agent use. Your past Claude chats stay in Claude. Start a project free

Sources

Checked on 17 October 2026.