MCP · OpenAI
Marketing MCP for ChatGPT
Add Spreeflo to ChatGPT as a custom MCP plugin and the chat you already plan in can build the segment, draft the email, and report on what it did.
Added from ChatGPT on the web·Sign in with OAuth·Asks before it writes
https://mcp.spreeflo.com/mcpWrites apply to your live workspace immediately — there's no approval queue here. Keep a human in the loop and check ChatGPT's confirmation settings first.
The agent
What is ChatGPT?
ChatGPT is OpenAI’s assistant. Its Plugins page can take a remote MCP server as a custom plugin, which is how it reaches Spreeflo — and because it reads each tool’s annotations, it knows which Spreeflo calls only look and which ones change something.
Called by name
Type @ and pick Spreeflo, and that message can use your workspace. Planning, copy and the live numbers sit in one conversation.
Confirms the writes
ChatGPT treats any tool not marked read-only as a write and asks before running it, so a segment or a draft shows up as a prompt in the chat first.
Workspace-ready
On Business and Enterprise, an admin can publish Spreeflo to the whole workspace and choose who gets it, instead of each person adding their own.
Setup
Connect ChatGPT in about a minute
One entry in ChatGPT, then sign in to Spreeflo. No key to create or paste.
Add the Spreeflo MCP server
Custom MCP plugin- 1.In ChatGPT on the web, open Plugins, click Add and choose Add custom MCP server.
- 2.Name it Spreeflo, paste the URL below, and pick OAuth. Leave the client ID and secret empty, acknowledge the risk warning, and create the plugin.
- 3.Install the new plugin. ChatGPT only offers a plugin in chats once it is installed.
https://mcp.spreeflo.com/mcpOn a Business or Enterprise workspace the Add option only appears once an admin has given you the “Create plugins with MCPs” permission.
Sign in and approve
Spreeflo opens in your browser. Sign in, pick the workspace ChatGPT should reach, and click Approve. Tick Make changes if it should be able to write — leave it off for a read-only connection.
Check it works
Ask ChatGPT for something small — "@Spreeflo list my audiences". A working connection is one where ChatGPT shows a Spreeflo tool call and answers with your real audience names.
Full setup guide and troubleshootingIn practice
What ChatGPT does with it
Two ordinary requests, and the Spreeflo tools behind them.
@Spreeflo turn these launch notes into a broadcast for customers on the annual plan.
Found the matching attribute, built the audience, and set up a draft broadcast — confirming each write with you first:
list_contact_attributescreate_segmentread_format_doccreate_email_templatecreate_broadcast
The broadcast is saved as a draft. Scheduling and sending stay in the Spreeflo app.
@Spreeflo how is the webinar sign-up form converting this week?
Found the form, then pulled its analytics and a summary of the responses:
list_formsget_form_analyticsget_form_summary
Reads only, so nothing stopped for a confirmation.
Running it
How ChatGPT decides when to stop and ask
ChatGPT turns Spreeflo’s tool annotations into prompts in the conversation, and on a company workspace an admin sits between the server and everyone else. Both shape how the connection behaves day to day.
Reads run, writes wait for you
Every Spreeflo tool that only reads is marked read-only, and ChatGPT treats everything else as a write. So listing segments or pulling campaign numbers goes straight through, while creating a segment or saving a template waits for your confirmation. An approval lasts for that conversation only. A new chat asks again.
Mention it in each message
Picking Spreeflo with @ applies to the message you send it with. Later in a long chat, a question without the mention may be answered from what ChatGPT already read rather than from a fresh call. Mention Spreeflo again when you want today’s numbers.
A published plugin is a snapshot
When an admin publishes Spreeflo to a Business or Enterprise workspace, ChatGPT keeps the tool list as it was at that moment. Tools Spreeflo adds later do not appear until the admin refreshes and republishes, and new actions arrive switched off until someone turns them on.
Deep research only reads
Deep research can call a custom MCP server, but only its read tools. That makes it a good way to write up a quarter of campaign performance, and it means a deep research run cannot create anything in your workspace.
What it reads can outlive the chat
With Memory on, ChatGPT may save details from a connected app to memory, and on personal plans app data can be used for training when “Improve the model” is on. Contact records are your customers’ personal data. Check those two settings before asking ChatGPT about individual contacts, or ask for counts and rates instead of names and emails.
Why this pairing
ChatGPT + Spreeflo
- Added from the Plugins page, nothing to install
- Signs in with your Spreeflo account — no key to paste
- Read-only tools are marked, so only writes ask first
- Admins can publish it to a whole ChatGPT workspace
- No MCP tool for sends, deletes or publishes
Coverage
Everything it can touch
The same surfaces you work with in the app, exposed as tools — and a short list that is deliberately off limits.
ChatGPT tends to work across the whole loop in one conversation: the numbers first, then a segment and a draft built from them, with a confirmation at each write.
Audience
- Audiences and contacts
- Segments, static and dynamic
- Tags
- Custom contact attributes
- CSV import
Sending
- Campaigns and journeys
- Email templates
- Web push templates
- Sender and domain identities
Capture
- Forms and question types
- Inline and popup embeds
- Response summaries
- CSV export of responses
Insight
- Campaign and web push performance
- Tracked events and pages
- Link clicks
- Form analytics
What MCP can’t do
These are irreversible or they reach real people, so they are removed from the surface rather than left to your client's confirmation settings. A person does them in the app.
Supercharge your marketing automation
FAQ
Questions, answered.
OpenAI’s help centre lists full custom MCP support for Business and Enterprise/Edu, and read-only custom MCP connections for Pro. A read-only connection still reports on everything; it just cannot create segments or drafts. If Plugins has no option to add a custom MCP server, check the plan first, then, on a company workspace, the “Create plugins with MCPs” permission. Nothing on the Spreeflo side changes with the ChatGPT plan.
No. OpenAI dropped that step: ChatGPT now shows a risk warning for each MCP server you add, and you acknowledge it before the plugin is created. Guides that start in Settings → Connectors → Advanced describe the older flow.
Nothing. ChatGPT registers itself with Spreeflo’s sign-in server when you connect, so those fields stay empty. The only things you supply are the server URL and your Spreeflo sign-in, and there is no API key to paste or rotate.
No. Spreeflo runs a hosted MCP server, so ChatGPT calls it directly over Streamable HTTP — nothing to install, nothing to self-host, no SDK to keep up to date. A server URL and a sign-in (or, for a headless agent, an API key in the Authorization header) are the whole integration.
It cannot delete anything or start a send. Spreeflo has no MCP tool for deletes, starting or resuming a campaign or journey, sending or scheduling a broadcast, or publishing a form, rather than leaving them to client settings. Those stay in the Spreeflo app, where a person does them. Journeys you have already switched on keep running, though, so editing a contact, its tags, a segment or an email template, or importing contacts, can set one off, just as the same edit would in the app.
The key resolves to exactly one workspace, and every action runs as the person who owns that key — their membership re-verified on each request. The audit trail names a real person, so agent-made changes are attributable.