❗ Heyflow's MCP is under active development and details might change. We will update this article accordingly.
This article walks through what the connector can do, how to set it up, and what data your assistant receives when you use it.
Once connected, you can ask your assistant questions about your funnels, look up help articles, and check analytics without leaving the conversation.
What is MCP?
The Model Context Protocol is an open standard for connecting AI assistants to external services. It works like a browser extension, but for AI: once you authorize the Heyflow connector in your assistant, it can pull information from your Heyflow flows on demand while you chat.
Heyflow's MCP server is currently read-only. Your assistant can answer questions about your funnels, but it cannot create, edit, publish, or delete anything – and it cannot read the individual responses people submit through your flows.
What can you ask?
Once the connector is set up, your assistant can help with three categories of questions.
1. Your Funnels
List the funnels you have access to in Heyflow and filter them by name.
"What funnels do I have?"
"Which of my funnels has the most responses?"
"Find my signup form."
For each funnel you'll see the ID, slug, title, response count, and whether it's published.
2. Funnel Analytics
Pull analytics for a specific funnel, including per-screen drop-off.
"How is my 'onboarding' flow performing?"
"What's the conversion rate for flow
abc123this month?""Where are users dropping off on my signup flow?"
"Show me analytics for the last 7 days."
You'll see total visits, total submissions, conversion rate, and a screen-by-screen breakdown with drop-off percentages.
3. Heyflow Documentation and How-to's
Search the Heyflow help center from inside your assistant.
"How do I add a file upload field to my flow?"
"What integrations does Heyflow support?"
"How do I set up webhooks?"
"Can I use conditional logic across multiple screens?"
The assistant answers with links back to the relevant help articles so you can read more if you need to.
Set up
The Heyflow MCP server uses two standard connection methods. Use the one your assistant supports.
HTTP + OAuth (recommended)
This is the full-featured option and unlocks all three tools (documentation, funnels, analytics).
Server URL:
https://mcp.heyflow.com/mcpTransport: HTTP
Authentication: OAuth 2.0
The exact setup steps depend on your assistant. Most MCP-compatible assistants have an Add MCP server option in their settings where you paste the URL and the client walks you through an OAuth login in your browser. Log in with your normal Heyflow credentials.
Once authorized, the assistant will show the Heyflow connector as connected and all three tools will be available.
Data sent when you connect
When you ask your assistant something that triggers a Heyflow tool, the connector sends the following to our servers:
Your identity. After you log in, your Heyflow user ID and email are included in the request so we can show you only your own funnels and analytics.
The question or filter. For example, your search query, the flow ID you asked about, or the date range.
A short log line for each tool call (tool name, user ID, result count) so our team can debug issues.
We do not send your assistant the full contents of your flows or individual responses. The assistant only sees what the tool returns in its reply, which is the same thing you'd see in the Heyflow app.
🔎 For the full details, including our third-party processors for this feature, see the Heyflow Privacy Policy.
Troubleshooting
The connector can't connect or the status is stuck on "needs authentication."
Open your assistant's MCP settings and re-authenticate the Heyflow entry. If that doesn't work, log out of Heyflow in your browser and try again.
You see "Access denied" for a specific flow.
You can only ask about funnels you own or have editor access to. Check that the flow ID belongs to your account.
💬 Questions, bug reports, or security concerns? Reach out to our team via the chat or [email protected]. Include the reference ID if your assistant showed you one. This helps us trace the request in our logs.