Where does the configuration go?
A developer-mode app in ChatGPT, signed in to nofeb with OAuth; from your own code, the OpenAI Responses API mcp tool with an API key.
How do I connect ChatGPT?
- In ChatGPT, open Settings, then Security and login, and turn on Developer mode (Plus, Pro, Business, Enterprise and Education accounts).
- Go to ChatGPT Plugins, select the plus button and create a developer-mode app with the MCP server URL https://api.nofeb.com/mcp.
- Choose OAuth for authentication and leave the static client credentials empty. Choose client ID metadata documents if offered; dynamic client registration also works.
- Connect, then in the nofeb console sign in, pick the organization and, to confine the app, one inbox, and approve the scopes. The app appears under Drafts in ChatGPT's app settings, where its tools can be turned on or off.
# ChatGPT app (developer mode): Settings > Security and login > Developer mode
# ChatGPT Plugins > + > developer-mode app
# MCP server URL: https://api.nofeb.com/mcp
# Authentication: OAuth (no static credentials; CIMD or DCR)
#
# nofeb's discovery documents, which ChatGPT reads for you:
# GET https://api.nofeb.com/.well-known/oauth-protected-resource/mcp
# GET https://api.nofeb.com/.well-known/oauth-authorization-server
# authorization_endpoint /oauth/authorize (PKCE S256 only)
# token_endpoint /oauth/token (refresh tokens rotate)
# registration_endpoint /oauth/register
# revocation_endpoint /oauth/revoke
# client_id_metadata_document_supported: true
# OpenAI Responses API, from your own code, with an API key:
resp = client.responses.create(
model="<model>",
input="List my nofeb inboxes",
tools=[{
"type": "mcp",
"server_label": "nofeb",
"server_url": "https://api.nofeb.com/mcp",
"headers": {"Authorization": f"Bearer {os.environ['NOFEB_API_KEY']}"},
"require_approval": "never",
}],
) Checked against ChatGPT's documentation on . source: developers.openai.com
What should I know about ChatGPT's setup?
- OpenAI's authentication guide says ChatGPT cannot present custom API keys to a connector, which is why the app signs in with OAuth. ChatGPT exchanges the code as a public client, which nofeb's token endpoint accepts with PKCE.
- Your role in the nofeb organization caps what you can approve: admins and owners can grant the full agent scopes, while a member can grant a third-party app read access to inboxes, metadata and calendars only. A tool that needs more gets 403 insufficient_scope naming the missing scope.
- Disconnect from the connected-apps list in the nofeb console, or with
DELETE /v1/oauth/grants/{grant_id}; tokens are checked on every request, so revocation takes effect immediately. - The Responses API mcp tool takes a headers map, which is how an API key reaches nofeb from your own code; OpenAI's servers make those calls, so nothing runs on your machine.
What can I ask it to do?
List my nofeb inboxes and summarise anything unread in the support inbox.
117 tools are available, including create_inbox, wait_for_email, reply_to_message, get_thread_context, set_thread_state and get_dns_setup. Whether a send goes out at once or waits for a person is decided by the inbox's autonomy tier on the server, not by ChatGPT.
Frequently asked questions
Can I connect nofeb to ChatGPT developer mode?
Yes, with OAuth. Create a developer-mode app for https://api.nofeb.com/mcp, choose OAuth, and approve the connection in the nofeb console. ChatGPT can identify itself with a client ID metadata document or register itself dynamically; nofeb supports both.
Can the ChatGPT app send mail as me without asking?
Only as far as the inbox's autonomy tier allows. A send from a send_with_approval inbox waits for a person, and the token cannot approve it unless whoever connected the app granted send:approve, which only admins and owners can.
Does the Responses API route need anything installed?
No. OpenAI's servers call nofeb's MCP endpoint directly from their side; nothing runs on your machine.
Status of everything on this page
Every capability named above, with its status as reconciled against the engine.
| Capability | What it does | Status |
|---|---|---|
| MCP server | A remote MCP server over streamable HTTP whose tools are generated from the same operation table as the REST API and dispatched through the same router, so a tool cannot behave differently from its endpoint. | Shipped |
| Scoped API keys | Keys bound to the organization, one pod or one inbox, holding only the scopes they are given. A key cannot mint a key wider than itself. | Shipped |
| Autonomy tiers enforced by the server | Every inbox carries a sending policy, from read-only to autonomous, enforced on the server for every surface that sends, so switching from the API to MCP cannot bypass it. | Shipped |
| Wait for an email | Long-poll an inbox for up to a minute until a message matching sender, subject or has-a-code arrives. | Shipped |
| TypeScript and Python SDKs | Official clients generated from the OpenAPI document: @nofeb/sdk for Node, Deno, Bun and edge runtimes, and nofeb for Python with sync and asyncio clients. Both chain long polls for waitForEmail, retry idempotent requests, page through lists and verify webhook signatures. | Shipped |
| MCP sign-in with OAuth | MCP clients connect by signing in instead of pasting a key: OAuth 2.1 with discovery, dynamic client registration, client ID metadata documents, PKCE, tokens bound to the MCP server, refresh-token rotation and immediate revocation. A person approves the scopes, and optionally one inbox, in the console, and can disconnect the app at any time. API keys keep working beside it. | Shipped |