--- title: "Connect Userflow to Claude, ChatGPT, and other AI tools with the Userflow MCP" slug: "mcp" updated: 2026-10-26T16:17:00Z published: 2026-10-26T16:17:00Z canonical: "help.userflow.com/mcp" --- > ## Documentation Index > Fetch the complete documentation index at: https://help.userflow.com/llms.txt > Use this file to discover all available pages before exploring further. # Connect Userflow to Claude, ChatGPT, and other AI tools with the Userflow MCP The Userflow MCP server connects your Userflow account to Claude, ChatGPT, Codex, and other AI tools. MCP stands for Model Context Protocol, an open standard that lets AI tools connect to other apps. Once it's connected, you can ask questions in plain language about your flows, users, segments, events, and Adoption Agent conversations, and have your AI tool build flows, segments, charts, and dashboards for you. > **Who can use the Userflow MCP:** everyone on your team. Each person signs in with their own Userflow login, so the Userflow MCP can only see and do what their Userflow role allows, and every action is credited to them. To change what someone can do, change their role in [Managing Teams in Userflow](https://help.userflow.com/docs/team-management), or revoke their connection (see [Managing your connections](/docs/mcp#managing-your-connections)). ## What you can ask the Userflow MCP Once the Userflow MCP is connected, you can ask questions in plain language about anything in your Userflow account. A few examples: ### Your flows - "How many flows do I have published, and which ones are still in draft?" - "When was the 'Welcome Tour' first published, and who's edited it most recently?" - "Which flows are tagged 'Onboarding' and are currently live?" ### Performance and analytics - "How many people saw the 'Product Tour' yesterday?" - "What's the completion rate for the 'Trial Signup' flow this month?" - "Compare completion rates between 'Onboarding A' and 'Onboarding B'." - "Summarize the NPS responses we received in the last 30 days." ### Your users and companies - "Did jane@acme.com complete the setup flow?" - "Show me all companies in the 'Trial' segment." - "Which users on the Free plan haven't logged in for 10 days?" ### Your Adoption Agent - "How many conversations did the Adoption Agent have last week?" - "What are the most common topics customers asked about?" ### Your account setup - "List all the events we're tracking." - "What user and company attributes are defined on this account?" ### Your help content - "What does our help center say about setting up single sign-on?" - "Find the help article about inviting team members." - "Summarize what our knowledge base covers about billing." The AI tool will pick the right data to look at based on your question. You can drill in, ask follow-ups, and combine filters the same way you'd talk to a teammate. ## What the Userflow MCP can build The Userflow MCP doesn't only answer questions. It can also build and change things in your Userflow account, so the flow or segment an answer points to can be created in the same conversation. A few examples: ### Build flows and campaigns - "Create a banner announcing Tuesday's maintenance window." - "Duplicate the 'Welcome Tour' and call it 'Welcome Tour: Enterprise'." - "Build an NPS survey for companies who have been customers for more than six months." ### Build segments - "Create a segment of companies on a trial that haven't published a flow yet." - "Segment the users who saw the onboarding flow but didn't finish it." ### Build charts and dashboards - "Build a funnel chart for activation over the last 30 days and add it to my activation dashboard." ### Publish a flow - "Publish the 'Trial Signup' flow." Creating or editing a flow always produces an unpublished draft, so nothing reaches your users until you say so. Publishing is a separate step that asks you to confirm every time. Segments, charts, and dashboards have no draft state in Userflow, so those are created in your account right away. For the full picture, including how to keep the MCP read-only if you'd rather, see [Creating and publishing with MCP write tools](https://help.userflow.com/docs/mcp-write-tools). The launch post, [Introducing MCP Write Tools](https://www.userflow.com/blog/blog-introducing-mcp-write-tools?utm_source=help.userflow.com&utm_medium=referral&utm_campaign=mcp-write-tools&utm_content=kb-mcp), covers why it's built this way. ## Which AI tools can connect The Userflow MCP works with AI tools that let you add a custom MCP server by URL. When you sign in, Userflow checks that the tool is on its list of approved MCP clients. Claude, ChatGPT, Codex, Cursor, Claude Code, Notion, and MintMCP all connect. The steps are the same everywhere: 1. Find where your tool lets you add a custom MCP server or connector. 2. Add `https://api.userflow.com/mcp` as the server URL. 3. Sign in with your Userflow login to authorize it. The Userflow MCP setup steps for Claude, ChatGPT, Cursor, Claude Code, and Codex are in [How to connect](/docs/mcp#how-to-connect). They aren't special cases: the same URL and the same sign-in work in any approved tool. If you can't find a way to add a custom MCP server in your tool, that tool doesn't support custom MCP connections yet, and that usually changes as vendors add MCP support. If a tool isn't approved yet, the Userflow sign-in fails with an error that says client_id is not an allowed CIMD URL. Contact support with the name of the tool, and we can look at adding it. ## How to connect Setup has two parts: adding the Userflow MCP as a custom connector in your AI tool, and signing in to authorize it. You can find Userflow in the ChatGPT connector directory. A native listing in Claude is coming. Until then, add Userflow to Claude as a custom connector, as described in [Add the Userflow MCP as a custom connector](/docs/mcp#1-add-the-userflow-mcp-as-a-custom-connector). Adding it by URL works in any AI tool with custom MCP support. ### 1. Add the Userflow MCP as a custom connector Use this server URL wherever your AI tool lets you add a custom MCP or connector: - `https://api.userflow.com/mcp` This one URL works for every account, in every region. If your account or some of your environments are hosted in the EU, you still use the same URL—you choose which environment to query within the conversation, and Userflow routes to the right region for you. **In Claude:** - **Personal account:** go to **Settings → Connectors**, click the **+** to add a custom connector, enter a name (for example, "Userflow MCP") and the server URL `https://api.userflow.com/mcp` in the dialog that opens, and click **Add**. Then complete the sign-in flow to finish connecting. - **Team or Enterprise account:** an admin adds the connector once, using the same steps as an individual account. After that, anyone on the account can open **Settings → Connectors**, find the listing the admin created (it has the name they gave it), select it, click **Connect**, and complete the sign-in flow with their own Userflow login. **In ChatGPT** and other AI tools, add the same server URL wherever the tool lets you add a custom MCP or connector. ### 2. Sign in to authorize When you click **Connect**, you're taken to a Userflow sign-in and consent screen in your browser. Sign in to Userflow the way you normally do, review the access you're granting, and approve it. You're redirected back to your AI tool, fully connected—there's no API key to create or paste. Because you sign in as yourself, the connection carries your own Userflow permissions: you can do exactly what you can do in the Userflow app, and nothing more. Once authorized, your AI tool can query your Userflow data. Try asking "How many flows do I have published?" to confirm the connection is working. > **Already connected during the early beta?** If you set up the MCP earlier using a Userflow API key, you'll be asked to sign in again the next time you use it. Follow the Userflow sign-in and consent screen to reconnect—this applies to EU accounts too. You don't need to remove or re-add the connector. ### Connecting from Cursor Cursor needs a pre-registered client rather than the standard sign-in, so use the Cursor install link instead of adding the server URL by hand. Adding `https://api.userflow.com/mcp` manually in Cursor returns an error about dynamic client registration. Open this link on the machine where Cursor is installed: ```plaintext cursor://anysphere.cursor-deeplink/mcp/install?name=Userflow%20MCP&config=eyJ1cmwiOiJodHRwczovL2FwaS51c2VyZmxvdy5jb20vbWNwIiwiYXV0aCI6eyJDTElFTlRfSUQiOiIyZTdmNmI4Mi0xYjNmLTRlNWEtOWMyMS02ZDhmMGIxYTJjMzMifX0= ``` Cursor opens and offers to install the Userflow MCP. Accept, then sign in with your Userflow login as usual. Your own permissions apply, exactly as they do in any other tool. ### Connecting from Claude Code Claude Code can connect to the Userflow MCP. In your terminal, add the server: ```plaintext claude mcp add --transport http userflow https://api.userflow.com/mcp ``` Then run `/mcp` inside Claude Code, select **userflow**, and sign in with your Userflow login in the browser window that opens. Your own Userflow permissions apply, the same as in any other tool. If you added the Userflow MCP to Claude Code before October 2026 and saw an error about an invalid redirect, remove the server and add it again. ### Connecting from Codex Codex connects to the Userflow MCP from the command line. In your terminal, add the server: ```plaintext codex mcp add userflow --url=https://api.userflow.com/mcp ``` Then sign in with your Userflow login to authorize it. Your own Userflow permissions apply, the same as in any other tool. ## Managing your connections You can see every active AI connection to your Userflow account—and revoke any of them at any time—from the connection management screen in your Userflow settings. Revoking a connection immediately stops that AI tool from querying your data until it's authorized again. ## What the Userflow MCP handles well, and what it can't do yet - **Structural and analytics questions both work.** Flow counts, status, ownership, segment membership, event definitions, and attributes, plus completion rates, drop-off, NPS summaries, and comparative analytics. Most data use cases are covered. - **Complex multi-step queries may need rephrasing.** If you don't get the answer you expect, try breaking the question into smaller parts. - **Some things still need the Userflow app.** Images need a public URL, page elements are selected in the Builder, and the content of checklists, launchers, and resource centers is added there too. See [What write tools can't do yet](https://help.userflow.com/docs/mcp-write-tools#what-write-tools-cant-do-yet). ## Troubleshooting ### "client_id is not an allowed CIMD URL" This error means the AI tool you're connecting from isn't on Userflow's list of approved MCP clients yet. It isn't a problem with your Userflow account, your permissions, or the server URL, and changing the redirect settings in your tool won't fix it. Claude, ChatGPT, Codex, Cursor, Claude Code, Notion, and MintMCP are approved. If you see this error in another tool, contact support with the tool's name. ### "Incompatible auth server: does not support dynamic client registration" You'll see this in Cursor if you added `https://api.userflow.com/mcp` as a custom server by hand. Cursor expects a pre-registered client. Use the install link in [Connecting from Cursor](/docs/mcp#connecting-from-cursor) instead. Adding a Bearer token doesn't change the outcome. ### Nothing happens after you sign in Check whether you connected during the early beta using a Userflow API key. Those connections need a fresh sign-in. Follow the sign-in and consent screen in [Sign in to authorize](/docs/mcp#2-sign-in-to-authorize) to reconnect, including on EU accounts. You don't need to remove or re-add the connector. ## Still need help with the Userflow MCP? If you run into trouble connecting or have questions about what the MCP can answer, reach out to your Userflow contact or post in the in-app help. ## Related articles - [MCP write tools: create flows, segments, and charts from your AI tool](https://help.userflow.com/docs/mcp-write-tools) - [AI Skills Library](https://help.userflow.com/docs/ai-skills-library): ready-made skills that use the Userflow MCP to report on your data and build things - [Managing Teams in Userflow](https://help.userflow.com/docs/team-management)