Use Ormind from the AI you already use

How it works

Ormind is an MCP server: a standard door that AI assistants know how to open. You give your assistant the address above, sign in with your Ormind account, and choose what it may do. From then on it works in your dossiers with your rights: what you can read, it can read; what you cannot, it cannot see. Every action is traced in the dossier journal in your name.

• Nothing to copy but the address and a Client ID. Claude and ChatGPT get their own credentials from the sign-in. No API key, no token, no secret.
• Two access levels. At sign-in you grant read only, or read and write. Read-only hides every writing tool from the assistant.
• Pages show as pages. In Claude, a wiki page or a deliverable opens as a panel inside the chat, with clickable links, not as a wall of text. Folders, search results, tables, the journal and your list of dossiers open as panels too.

Claude

Works on claude.ai, the desktop app and the mobile apps, on every plan (Free plans allow one custom connector).

1. Open the connectors pageClick your name, then Settings › Connectors (on some accounts the menu is called Customize › Connectors).
2. Add a custom connectorClick +, then Add custom connector. Name it Ormind and paste the address https://app.ormind.ai/api/mcp.
3. Give it the Client IDOpen Advanced settings on the same form and paste this into OAuth Client ID (leave Client Secret empty), then click Add:
   OAuth Client ID

   loading…Copy
   Without it, Claude registers a new sign-in client on every connection, and our sign-in service caps how many it keeps.
4. Connect and sign inClick Connect next to Ormind. Sign in with your Ormind account, then choose read only or read and write, and approve.
5. Turn it on in a chatIn a conversation, click + › Connectors and switch Ormind on. Then ask: “List my workspaces.”
6. Allow the panelThe first time Claude opens a page, it asks permission to display the Ormind app. Choose Always allow. Pages and deliverables then render inline.

Tip: in the connector's settings you can set the reading tools to Always allow, so Claude stops asking before each read.

ChatGPT

Custom connectors are available on Plus, Pro, Business, Enterprise and Edu plans, on the web, behind a switch called Developer mode.

1. Turn on Developer modeClick your name, Settings › Connectors (also shown as Apps & Connectors) › Advanced settings, and switch on Developer mode.
2. Create the connectorBack on Connectors, click Create. Name: Ormind. MCP server URL: https://app.ormind.ai/api/mcp. Authentication: OAuth.
3. Give it the ChatGPT Client IDUnder the OAuth settings, set the registration method to User-defined OAuth client and paste this into OAuth Client ID (leave the secret empty). The callback URL ChatGPT shows, https://chatgpt.com/connector_platform_oauth_redirect, is already allowed. Tick I trust this application, then Create:
   OAuth Client ID for ChatGPT

   loading…Copy
   Do not let ChatGPT register a client automatically: our sign-in service caps how many it keeps, and the connection would fail.
4. Sign inChatGPT sends you to the Ormind sign-in page. Sign in, choose read only or read and write, approve. ChatGPT then lists the available actions.
5. Use it in a chatIn a new conversation, click + › More › Developer mode and select Ormind. Then ask: “List my workspaces.”

Claude Code

For the terminal, Ormind gives you a personal agent token instead of a sign-in screen. The token is shown once; regenerating it revokes the previous one.

1. Get your tokenOpen your connections in Ormind (also reachable from any dossier's overview, "Work on this dossier from your own AI" › Your token), then, under With Claude Code on your computer, generate your agent token.
2. Paste the commandOrmind shows you the exact command. It looks like this:

   claude mcp add --transport http ormind https://app.ormind.ai/api/mcp \
     --header "Authorization: Bearer cg_…your token…"

   Prefer a file? The same page gives you a ready-made .mcp.json to drop in your project.
3. CheckIn Claude Code, type /mcp: Ormind should be listed as connected, with its tools.

Any other MCP client that speaks HTTP (Codex, Cursor, VS Code…) can use the same address and the same Authorization: Bearer header.

What you can ask

Name the dossier the first time; the assistant remembers it for the conversation.

Rights, traces and revocation

• Your rights, exactly. The assistant holds your role in each workspace: reader, editor or owner. A refusal is final and the assistant is told so.
• Every call is journaled in the dossier, in your name, with the assistant's name (for example Serge's Claude).
• Revoke any time. Claude: Settings › Connectors › remove. ChatGPT: Settings › Connectors › delete. Claude Code: regenerate or delete the agent token in your connections.

Troubleshooting