Option 1 — install the Rings app (fastest)
Rings is published in the ChatGPT app directory. Nothing to configure.Install Rings AI in ChatGPT
Open the listing, click Connect, and sign in with your Rings account.
Option 2 — add it yourself in developer mode
Use this when you want the full tool surface, including write actions, under your own workspace’s controls.Developer mode and full MCP support are available on ChatGPT Business,
Enterprise, and Edu plans, on ChatGPT web. On Business, only admins and
owners can enable it; on Enterprise and Edu, admins can grant it to specific
members via RBAC.
1
Enable developer mode
Admins and owners turn it on in Workspace Settings → Permissions &
Roles → Connected Data → Developer mode / Create custom MCP connectors.Each person then enables it for their own account under Settings → Apps →
Advanced Settings.
2
Create the app
Go to Settings → Apps → Create (admins can also do this from
Workspace settings → Apps → Create).
3
Fill in the server details
4
Scan tools
Click Scan Tools. Complete the Rings OAuth prompt when it appears, then
wait for the scan to list the tools.
5
Create
Click Create. The app appears under Settings → Apps → Enabled Apps
with a Dev label, and as a draft in Workspace settings → Apps →
Drafts.
6
Publish it to the workspace
An admin or owner opens Workspace settings → Apps → Drafts and clicks
Publish. Enterprise and Edu admins can use Configure Actions to
choose which tools are allowed and Configure Access to pick which groups
get the app.
ChatGPT loses access when the first token expires unless the server issues
refresh tokens. Rings advertises the
refresh_token grant in its
authorization-server metadata and issues refresh tokens, so no extra
configuration is needed on your side.Option 3 — the OpenAI API
Point the Responses API straight at the Rings MCP server:authorizationtakes an OAuth access token. OpenAI does not store it, so send it on every request.- Narrow the surface with
allowed_tools: ["tool_a", "tool_b"]— get the exact names from an MCPlist_toolsrequest. - Leave
require_approvalat its default when the model can call write tools.