Connect your assistant

Add Rindo to the Claude, ChatGPT or Grok you already use, or hand Claude Code a personal access token.

Last updated 2026-10-12

Rindo is an MCP server at https://app.rindo.io/mcp. Connect the assistant your team already uses and it works as you: your projects, your permission, never more. With Claude, ChatGPT and Grok it is read-only until you grant write, and there is nothing to install or paste.

Claude — web, desktop and mobile

1. In Claude, open Settings → Connectors and add a custom connector with the URL https://app.rindo.io/mcp.
2. Claude opens Rindo in your browser. Sign in, then approve the connection. Read only is pre-selected; choose Read and write only if you want the assistant to create and edit tasks and pages for you.
3. That is all. The connection covers every project you are a member of — now and later — at the permission you hold in each.

The same connector follows your Claude account to the desktop and mobile apps. You can see and revoke every connected app in Rindo under Account settings → MCP access → Connected apps.

By default Claude asks before every tool runs, reads included. Open the connector from Settings → Connectors on claude.ai: its page groups the tools as Read-only tools and Write/delete tools, with a blanket permission per group and one per tool. Set the read-only group to Always allow and reads run without asking, while every write keeps asking until you allow it.

Roles and project guidance

• If a project has given you a work role (BA, QA, reviewer…), the connector lists it in Claude’s prompt picker. Pick it and the role’s instructions attach to your next message. A role assigned after you connected appears after Refresh tools list in the connector’s menu.
• A project’s guidance can be attached to a message from + → Connectors → Add from your Rindo connector. A project you joined after connecting shows up after you disconnect and connect again.
• To bring a document into the conversation, ask the assistant to read it by name — it has the tools to find and open it.

ChatGPT

ChatGPT adds Rindo as a plugin, on a paid ChatGPT plan (tested on Pro):

1. Open Plugins in ChatGPT, choose Add, then Add custom MCP server. Give it a name, set the Server URL to https://app.rindo.io/mcp, leave Authentication on OAuth, tick the risk notice and choose Create as a plugin. Nothing else to enter: ChatGPT discovers Rindo’s sign-in on its own.
2. ChatGPT offers Continue to Rindo. Sign in, then approve the connection — Read only is pre-selected; choose Read and write to let it create and edit for you.
3. In a chat, type @ and pick Rindo, then ask. Reads run without asking; before each change ChatGPT asks Allow ChatGPT to use Rindo? — allow once, or always.

The connection covers every project you are a member of, at the permission you hold in each: where you are a Viewer, a write is refused even with a write grant. ChatGPT appears in Rindo under Account settings → MCP access → Connected apps like any other connected app.

Grok

Grok adds Rindo as a custom connector on grok.com (tested 2026-10-12):

1. Open Plugins in Grok, choose New Connector, then Custom. Give it a name, set the Server URL to https://app.rindo.io/mcp and add it. Nothing else to enter.
2. Grok opens Rindo in a new window. Sign in, then approve the connection. The consent screen says Unverified app — it registered itself: that is expected, because Grok registers itself with Rindo each time it connects instead of presenting an identity Rindo can check; continue only if you started the connection yourself. Read only is pre-selected; choose Read and write to let Grok create and edit for you.
3. Ask Grok, naming the project. Grok does not ask before a tool runs — reads, creates and edits all run as soon as you ask — so the access you approved is the only limit you set. Keep Read only unless you want it to write.

The connection covers every project you are a member of, at the permission you hold in each. Reconnecting (Reauthenticate in the connector’s details) registers Grok again and shows the same notice. Grok appears in Rindo under Account settings → MCP access → Connected apps, once per connection; revoke the ones you no longer use there.

Claude Code

Claude Code connects with a personal access token instead of a sign-in:

1. In Rindo, open Account settings → MCP access → Tokens and create a token. Decide what it may do before you hand it to an agent — make it read-only, or narrow it to the capabilities you choose — and copy it: it is shown once.
2. Register the server:

claude mcp add --scope user --transport http rindo https://app.rindo.io/mcp \
  --header "Authorization: Bearer rnd_…"

The token is your own key. Revoke it from the same page at any time; the agent loses access on its next call. In Claude Code, any Rindo document can also be read by its address: ask the agent to read rindo://p/<project key>/documents/<id> — the key and id are in the document’s link.

Other MCP clients

Other MCP clients connect the same way as Claude Code, with a personal access token.

What your assistant can and cannot do

• It has no delete tools, no settings tools and no membership tools. Over MCP those actions are absent, not denied.
• Every write it makes lands on the project’s audit trail, with your name and the channel it came through.
• Pages are versioned: any edit can be compared and rolled back.
• Flip one project switch — Settings → General → Agent writes, “Content writes via MCP require approval” — and every edit made over MCP to an existing page lands as a pending draft with a diff until it is approved. Approval is a separate act, and one your own assistant can perform as you if its grant allows write. Approval is refused if the page changed underneath.