Can I share Claude’s memory with my team?

Partly. A project’s CLAUDE.md can be shared through source control, but it stays with that repository and Claude Chat does not read it. Claude Code’s auto memory stays on one machine, and Claude Chat memory belongs to one account. With CommonGround, every teammate’s Claude reads one shared wiki, in Claude Code and Claude Chat.

Can non-developers use the team wiki in Claude Chat?

Yes. They connect Claude Chat to the wiki with a browser sign-in, no API key. On Claude’s Team and Enterprise plans, a workspace Owner adds the CommonGround connector once, then members connect; on an individual plan, each person adds it. One approval covers every wiki that person belongs to, and Claude Chat answers from the one they choose as the default. Seeding happens in Claude Code.

How does a team knowledge base stay up to date?

Curators record things as they happen: in Claude Code, the ingest command turns a thought, notes, a transcript, a doc or a link into wiki pages. The lint command reports stale or orphan pages and broken citations, and changes nothing.

Is there a free plan for teams?

Not for team wikis. The Free plan is one personal wiki, a local folder Claude Code reads. Team wikis, with invites and roles, are on CommonGround Pro, which also puts your own wiki in Claude Chat and on your phone. During the private beta, request an invite to create a wiki; teammates join through an admin’s link.

Is the team wiki locked into CommonGround?

No. It’s plain markdown, and every member can keep a local copy as a folder that opens in Obsidian. If you leave, one clone takes the whole wiki and its history with you.

Team knowledge base for Claude

Every teammate’s Claude starts from the same page.

CommonGround is a team knowledge base: one wiki that every teammate’s Claude reads, in Claude Code and in Claude Chat. Record a decision once, and whoever asks next gets that answer and the page it came from.

One deploy rule, recorded once, then answered for a developer in Claude Code and a designer in Claude Chat.

Without CommonGround

Ravi wrote the rule into his project’s file. Wiley’s Claude, in another repo, never saw it. Mia, in chat, has no file at all.

Ravi · api/ · claude code

Added to api/CLAUDE.md, line 212: deploys go through staging first.

Wiley · web/ · claude code

Can we push this to production?

Yes. Merging to main deploys straight to prod.

Mia · designer · Claude Chat

How do we ship changes to the site?

I don’t have that context. Can you paste the deploy doc?

With CommonGround

Ravi records it once. Wiley’s Claude and Mia’s Claude read the same page and cite it, each in their own surface.

Ravi · claude code

/commonground:ingest deploys pass through staging first, always.→ decisions/deploys-pass-staging.md

Wiley · web/ · claude code

Can we push this to production?

Through staging first. The team decided this on August 14.decisions/deploys-pass-staging.md

Mia · designer · Claude Chat

How do we ship changes to the site?

Through staging first, then production. Here is the runbook.decisions/deploys-pass-staging.md · runbooks/deploy.md

One answer for the whole team, whoever asks.

Claude already remembers things, but none of that memory reaches every teammate on every surface. Two people can ask the same question and get different answers.

In Claude Code, a project’s CLAUDE.md can be shared through source control. It stays in that one project, and Anthropic suggests keeping it under 200 lines.
Claude Code’s auto memory stays on the machine that wrote it.
Claude Chat memory belongs to each person’s account.

One team knowledge base, read in Claude Code and Claude Chat.

Claude Code finds the wiki through a short pointer in the CLAUDE.md of each project you point at it; Claude Chat finds it through a connector. Claude reads the wiki’s index first, then only the pages a question needs, and names them in its answer.

A developer in the terminal, a product manager in the Claude desktop app and a designer on their phone all get the same answer. A charter says what belongs in the wiki and what stays out, and Claude checks it before it files anything new.

Everyone reads. Curators write. The history keeps every change.

In Claude Code, the plugin asks before anything reaches the shared wiki. Each change goes into the history with a line saying why it was made, so anyone on the team can look up when deploys started going through staging, and the reason.

Members read, search and ask. Publishing isn’t part of their role.
Curators record decisions, edit pages and publish them.
Admins curate too, invite people, and can revoke a Claude connector approved for the wiki. Access ends on its next request.

Started in one conversation, shared with a link.

Whoever creates the wiki in the web app becomes its admin and fills it from Claude Code. Seeding starts with the charter, then builds pages from a short interview, a folder or Obsidian vault, or documents you hand it. Stop partway and it picks up where you left off.

Then invite the team from the web app. The link creates each person’s account and puts them on the wiki with their role. Developers add the plugin in Claude Code; everyone else connects Claude Chat.

Keep your tools. Add one wiki every Claude reads.

If your team shares context one of these ways today, here’s what changes.

A project’s CLAUDE.md: Reaches Claude Code in one project. The wiki reaches Claude Chat and every project you point at it, and can keep growing, because Claude opens only the pages a question needs.
Claude Projects: A project keeps knowledge and instructions for its chats, shareable on Claude’s Team and Enterprise plans. The wiki serves Claude Code and Claude Chat alike, and a curator adds a decision by telling Claude.
Memory plugins and MCP memory servers: Claude reads pages a curator chose to publish: the same pages for the whole team, each with its history.
A Notion connector: Your docs stay where they are. Claude answers from focused wiki pages, and in Claude Code it can pull a Notion page into the wiki through your own connector.

What teams ask before sharing a knowledge base.

Can I share Claude’s memory with my team?: Partly. A project’s CLAUDE.md can be shared through source control, but it stays with that repository and Claude Chat does not read it. Claude Code’s auto memory stays on one machine, and Claude Chat memory belongs to one account. With CommonGround, every teammate’s Claude reads one shared wiki, in Claude Code and Claude Chat.
Can non-developers use the team wiki in Claude Chat?: Yes. They connect Claude Chat to the wiki with a browser sign-in, no API key. On Claude’s Team and Enterprise plans, a workspace Owner adds the CommonGround connector once, then members connect; on an individual plan, each person adds it. One approval covers every wiki that person belongs to, and Claude Chat answers from the one they choose as the default. Seeding happens in Claude Code.
How does a team knowledge base stay up to date?: Curators record things as they happen: in Claude Code, the ingest command turns a thought, notes, a transcript, a doc or a link into wiki pages. The lint command reports stale or orphan pages and broken citations, and changes nothing.
Is there a free plan for teams?: Not for team wikis. The Free plan is one personal wiki, a local folder Claude Code reads. Team wikis, with invites and roles, are on CommonGround Pro, which also puts your own wiki in Claude Chat and on your phone. During the private beta, request an invite to create a wiki; teammates join through an admin’s link.
Is the team wiki locked into CommonGround?: No. It’s plain markdown, and every member can keep a local copy as a folder that opens in Obsidian. If you leave, one clone takes the whole wiki and its history with you.