To configure an MCP server in ChatGPT, enable Developer mode, create a custom app in Settings → Apps, enter the server’s remote endpoint and metadata, choose authentication, scan its tools, and save a draft. Test the draft in a new chat; a workspace admin or owner must publish it for workspace use. Your plan, role, workspace policy, and current feature rollout determine which controls you can access. This guide covers the ChatGPT web setup, with a separate note on API configuration so the two paths are not confused. OpenAI’s Developer mode and MCP apps guide was updated in August 2026; interface labels and beta availability may change.
What you are configuring
In ChatGPT, an MCP server is connected through a custom app created in Developer mode. The app makes the server’s tools available in chats; it is not the same as adding a remote MCP tool to an OpenAI API request. ChatGPT connects to remote servers, so a server available only on your computer or private network needs a supported connectivity route, such as Secure MCP Tunnel.
The basic sequence is: confirm access, enable Developer mode, create the app, authenticate and scan tools, save a draft, test it, then have an authorized workspace admin or owner publish it.
Check plan, role, and workspace access
Availability depends on plan and administrator settings. OpenAI’s help article describes custom apps and Developer mode on ChatGPT web for Business and Enterprise/Edu. Business admins or owners handle workspace deployment. Enterprise/Edu admins can grant developer access through workspace role controls, after which authorized members enable Developer mode in user settings. The same article says Pro users can connect MCPs with read/fetch permissions in Developer mode; full MCP support, including write/modify actions, is rolling out in beta to Business and Enterprise/Edu. What you see can therefore differ by workspace and rollout status.
#1 Best Overall
If you cannot find Developer mode or Create, ask your workspace administrator whether the feature is enabled and whether your role may create apps. Do not assume that having a ChatGPT account alone grants permission to publish a workspace app.
Configure the MCP app in ChatGPT
- Enable Developer mode. In a workspace, an admin first enables the feature in workspace settings. Business users follow the Apps or Developer mode route available to their role in workspace or user settings. For Enterprise/Edu, an admin grants developer access and eligible members enable Developer mode in user settings. The exact navigation labels can vary by plan and may change.
- Open app creation. In the relevant Workspace or user Settings, go to Apps → Create. Confirm Developer mode is enabled.
- Enter the server details. Provide the MCP server endpoint and the required app metadata. Use the endpoint supplied by the server operator; ChatGPT must be able to reach it remotely.
- Choose authentication. Select the method required by the server. If the server uses OAuth, complete the authorization prompt when it appears.
- Scan tools. Choose Scan Tools and wait for discovery to finish. Review the tools ChatGPT finds before saving, especially any that can modify or send data.
- Save a draft. Select Create. The app is now a draft, not yet a published workspace app.
Check OAuth refresh behavior
For OpenID Connect, verify that the identity provider’s discovery metadata advertises the relevant endpoints and that it issues refresh tokens. OpenAI’s help instructions identify offline_access as the standard scope to request a refresh token. Without a refresh token or equivalent renewal mechanism, an initial authorization may expire and users may need to authenticate again.
Rank #2
Test the draft and publish it
Test in a new chat
Start a new chat, open the tools menu, and select the draft app, or refer to the app in your prompt. Try realistic requests that exercise each tool you expect people to use. Confirm that results are sensible and that any read, write, or modify action behaves as intended. Review app permissions before testing: depending on permissions and context, ChatGPT may ask for confirmation before a write action, and some high-risk actions may be blocked.
Publish through workspace settings
For workspace availability, an authorized admin or owner publishes the draft from Workspace settings → Apps → Drafts. Enterprise/Edu admins can manage who has access and set action permissions; they can also refresh actions to review tool changes. OpenAI says the workspace uses a frozen snapshot of available tools and inputs after initial approval until an admin reviews and publishes updates. Its help article also states that, at the time it was updated, Business apps could not be updated after publication and had to be recreated and republished. That plan-specific behavior is volatile, so check the live help article before relying on it.
Do these 3 things before closing this tab:
1Repair Windows errors before they cause bigger problems2Fix the driver behind crashes, sound loss and screen glitches3Clear out junk files and repair common Windows errorsUse the app in a chat
Start a chat and select the app for the message, or mention the app in your prompt. App selection applies to the message using it. If a later turn needs fresh data or another action, select or mention the app again.
Can I connect to a local MCP server?
Not directly. ChatGPT connects to remote MCP servers and cannot connect straight to a server running only on a local machine. For a private-network, on-premises, or developer-machine server, OpenAI points to Secure MCP Tunnel as a way to connect supported OpenAI products without exposing the server to the public internet. Another possible deployment choice is a publicly reachable hosted endpoint; OpenAI does not require a particular hosting provider. Confirm the current tunnel availability and setup requirements in the OpenAI setup guide.
Rank #4
Security and data handling before you connect
- Trust the operator. Connect only servers whose operator and behavior you trust. OpenAI warns that unsafe or untrusted servers can create security risks, including prompt injection.
- Review tool permissions. Know which tools can read information and which can create, send, or modify it. Limit access to users who should authorize those actions.
- Check the server’s data terms. Data sent to a remote MCP server is handled under that third party’s retention and residency policies. Review the specific operator’s terms before sending sensitive information. See OpenAI’s platform data-controls documentation.
- Revisit changes. Review newly discovered tools and inputs before publishing updates; do not assume a server’s tool list remains unchanged.
Are search and fetch tools required?
No. OpenAI’s help article says search and fetch tools are no longer required for connected servers. You can expose the tools appropriate to your server’s purpose, subject to your plan and workspace controls.
If you meant the OpenAI API
API-side MCP setup is a different configuration path: you define a remote MCP tool in an API request rather than creating a ChatGPT app in Settings. The tool schema requires either server_url or connector_id, plus a server label. Optional settings include authorization, headers, allowed_tools, and require_approval. Your calling application handles OAuth and supplies the access token. The API reference lists connector identifiers for Dropbox, Gmail, Google Calendar, Google Drive, Microsoft Teams, Outlook Calendar, Outlook Email, and SharePoint. Check the OpenAI API reference for the current request schema; these fields are not ChatGPT UI steps.
Common setup problems
| Symptom | Likely cause | What to check |
|---|---|---|
| Developer mode or Create is missing | The workspace has not enabled the feature, your role lacks access, or the feature is not available on your plan or current rollout. | Ask an admin to verify workspace settings and role access. Confirm that you are using ChatGPT web and the appropriate workspace. |
| Tool scan fails or finds no tools | The endpoint is unreachable, authentication is incomplete, or the server is not presenting tools as expected. | Check the endpoint with the server operator, confirm remote reachability and authentication, then run Scan Tools again. |
| OAuth works once, then access expires | The provider may not issue a refresh token or the request may omit the required refresh scope. | For OpenID Connect, check provider discovery and token behavior; ask whether offline_access is supported and requested. |
| A local or private server cannot be reached | ChatGPT cannot connect directly to a local-only service. | Use a supported route such as Secure MCP Tunnel or deploy a remotely reachable endpoint permitted by your security policy. |
| Draft works but colleagues cannot use the app | A draft is not published, or workspace access and action permissions have not been granted. | Have an authorized admin or owner review and publish it in Workspace settings → Apps → Drafts, then check access controls. |
| Tool behavior differs after a server change | The workspace may still use its approved snapshot of tools and inputs. | Ask an admin to refresh and review the actions, then publish the update according to the workspace’s current process. |
Or skip the browser setup
If what you need is a screenshot tool for an MCP-enabled workflow, ScreenshotNeo provides a screenshot API and MCP server for AI agents. One GET request can return a PNG, JPEG, WebP, or PDF. For an MCP client such as Claude or Cursor, its server includes take_screenshot, get_page_info, and capture_pdf. Cookie banners are accepted like a visitor and more than 60 known consent platforms, newsletter popups, and chat widgets are removed before capture; each cleanup step can be turned off. Bot checks, blank pages, timeouts, failed loads, and cache hits are not billed, with response headers identifying the page verdict and billing status.
Example cURL request (replace the URL with the page you want to capture):
curl -G "https://api.screenshotneo.com/v1/shot" -d access_key=YOUR_API_KEY --data-urlencode url=https://stripe.com -o shot.webpSee the ScreenshotNeo API documentation for API options and response details. It includes full-page capture, element capture by CSS selector, dark mode, device and viewport settings, retina scale, PDF page settings, custom CSS and JavaScript, selector waits, request blocking, custom headers and cookies, timezone and geolocation, caching, signed image links, async jobs, bulk capture, and a usage API. Every feature is on every plan. The Free plan includes 1,000 screenshots per month without a card; paid plans start at $5 for 3,000. Sign up free for 1,000 screenshots a month with no card.
Frequently Asked Questions
Do I need to add search or fetch tools to my MCP server for ChatGPT?
No. OpenAI says search and fetch tools are no longer required for connected servers.
Recommended Free Tools
Does creating a draft make the app available to my whole workspace?
No. An authorized workspace admin or owner must publish the draft, and workspace access controls determine who can use it.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
