# Org settings Feature toggles, MCP access, and what an org admin can change versus what is platform-locked. On the Organisation Settings page, at `/settings/org/settings`, you turn features on and off for everyone in your org. There is no Save button: each toggle takes effect when you click it. ## Layout The page has these cards: 1. **Chat & AI Capabilities**: what users can do inside a conversation 2. **Memory**: whether members can save facts from conversations 3. **Sharing & Collaboration**: member sharing and visibility 4. **Shell & MCP Access**: MCP servers and custom connectors 5. **Tool Call Approvals**: how strictly tool calls made by the AI are reviewed before they run 6. **Workspaces**: features members use in their workspaces 7. **Member Data**: what members can take out of PrivateMind themselves 8. **User Management**: shortcuts into the Members and Teams pages (documented separately in [Users and teams](https://docs.privatemind.com/users-and-teams.html)) Each toggle shows whether it is `On` or `Off`, with an Enable or Disable button. Some toggles carry a badge or a lock icon: see [Platform overrides](https://docs.privatemind.com/org-settings.html#platform-overrides). ## Chat & AI Capabilities | Toggle | What it controls | | ---------------- | ------------------------------------------------------------------------------------------------------ | | Web Search | Whether users can ask the model to search the web for real-time information during a conversation | | File Attachments | Whether users can attach PDFs, images, and other documents to messages for the model to analyse | | Code Execution | Whether the model is allowed to write and run code inside sandboxes to complete a task | | Skills | Whether the model can use reusable skills (document creation and more) automatically. Requires Code Execution. Sub-toggles govern whether members can create, share, and promote their own skills | | Automations | Whether users can dispatch long-running autonomous agents and build automated workflows | | Image Generation | Whether the model can create images from a prompt | | Voice Mode | Whether members can talk to the assistant by voice | | Account Lookup | Whether the assistant may answer a member's questions about their own account: settings, API keys, automations and connectors. Read-only, never secrets | These are **on** by default for a new org. Turning one off hides its control in chat and refuses the matching API call. For example, turning off Web Search removes the globe button from the composer. ## Shell & MCP Access | Toggle | What it controls | | ------------------ | --------------------------------------------------------------------------------------------------------- | | MCP Servers | The master switch for Model Context Protocol tools and connectors in your org | | Custom MCP Servers | Whether org admins can register their own external connectors: MCP servers and REST APIs. Opens the **Org Connectors** page (see below) | > [!NOTE] > **Code Execution**, in the Chat & AI Capabilities card above, switches sandboxes on and off. Turning it off stops the model running code or shell commands from the next message onwards, and Skills stop working with it. To add your own connectors, turn on both **MCP Servers** and **Custom MCP Servers**, then register each one on the [Org connectors](https://docs.privatemind.com/mcp-servers.html) page, as described in [Registering custom connectors](https://docs.privatemind.com/org-settings.html#registering-custom-connectors). ## Workspaces | Toggle | What it controls | | ----------- | ---------------- | | Secrets | Whether members can store passwords and API keys and attach them to their workspaces and jobs. PrivateMind turns it on for your organisation, and while it is off the switch is locked. An org admin can turn it off, which removes every secret's file from workspaces and keeps the secrets and their attachments. Terminals and notebook kernels that already loaded a secret as a variable keep it until they restart. See [Secrets](https://docs.privatemind.com/secrets.html#turning-secrets-on-and-off) | | App Central | Whether members can publish an app running in their workspace and choose who can open it. It stays off until PrivateMind turns it on for your organisation; an org admin can turn it off. Turning it off refuses new visits to every published app within about half a minute and cuts the live connection of pages already open within about a minute. It deletes nothing. See [App Central](https://docs.privatemind.com/app-central.html) | ## API access by IP The **API access by IP** card holds your organization's allowlist and blocklist of IP addresses. They cover every user API key and application key in your org. How they combine with a key's own rules, and what a refused request returns, is in [Restricting keys by IP address](https://docs.privatemind.com/api-keys.html#restricting-keys-by-ip-address). ## Platform overrides Some toggles can be **disabled platform-wide** by the PrivateMind platform team. When that happens the toggle: - shows in a dimmed, locked state - displays the message "Disabled platform-wide by administrator" - refuses the Enable click You cannot re-enable a platform-overridden feature from this page. If a feature you need is locked off, contact support. The override is one-way: platform admins can disable but not force-enable. Your local Off state still wins over a platform-wide On state. ## Registering custom connectors Your own connectors are external MCP servers and REST APIs. You register them on the Org Connectors page at `/settings/org/org-connectors` (**Settings, Org Connectors** in the sidebar). With **MCP Servers** and **Custom MCP Servers** on, an org admin adds each connector there by name and URL. PrivateMind then opens the egress path for it: the outbound network route to that connector's host. Registered connectors then appear in the **Connectors Library** for everyone in your org. Members can connect one, which adds it to **My Connectors**, and attach it in a chat. Registration is self-service for org admins. A member's request for a connector waits in the **Approval queue** on the Org Connectors page until an org admin approves or declines it. The full walkthrough is in the [Org connectors](https://docs.privatemind.com/mcp-servers.html) guide. > [!WARNING] > An external connector receives whatever a user sends it during a conversation. Register only services your organisation trusts. Each connector can reach only its own host. PrivateMind treats it as untrusted external egress, so it is dropped from any chat that also holds a private, internal source. ## What is **not** on this page These are platform-level controls, so an org admin can't change them here: - **Org name, display name, description, logo.** These come from your subscription record. Your PrivateMind account contact edits them, outside the app. - **SSO configuration.** Allowed email domains, IdP endpoints, and SAML metadata are configured outside the app, at the identity provider and platform layer. - **Sign-in security.** Your organisation's members must use multi-factor authentication. PrivateMind sets this policy. - **Data residency.** Your data is stored where your org was provisioned. Changing that means re-provisioning, which support handles. - **Subscription details.** Tier, billing cycle, user limit, storage limit, and feature access show on the Overview page. The PrivateMind platform team manages them. ## Where next - [Users and teams](https://docs.privatemind.com/users-and-teams.html): manage members and team groupings - [Audit logs](https://docs.privatemind.com/audit-logs.html): every toggle flip is logged with who and when - [Budgets and limits](https://docs.privatemind.com/budgets-and-limits.html): metering and spend controls for the org - [Secrets](https://docs.privatemind.com/secrets.html): what the Secrets switch turns on, and what turning it off does