The Appcues MCP Server connects AI tools like Claude, ChatGPT, and Cursor to your Appcues account so you can build experiences (beta), query experiences, analyze data, and manage workflows using natural language instead of navigating the dashboard.

What the MCP Server does

MCP (Model Context Protocol) is an open standard that lets AI assistants interact with external tools and data. The Appcues MCP Server uses this protocol to give your AI client secure access to your Appcues account — Flows, experiences, segments, goals, NPS data, events, and more. 

You work with your AI client using natural-language requests. That client uses the MCP Server to make the right Appcues tool calls, fetch the data, and return it in a conversational format.

Prerequisites

• An active Appcues account with admin access (to enable the MCP Server)
• An MCP-compatible AI client (Claude, ChatGPT, Cursor, or similar)

Set up the MCP Server

Go to Settings > Appcues AI. You can go directly to app.appcues.com/settings/ai.

Toggle the MCP Server setting to On. This enables MCP access for your account.

Open your AI client and register the Appcues MCP Server URL:

https://mcp.appcues.com/mcp

For accounts on Appcues' EU data hosting, register the Appcues MCP Server URL:

https://mcp.eu.appcues.com/mcp

Which URL should I use? Most Appcues accounts use the standard URL (mcp.appcues.com/mcp), regardless of where your company is located. The EU URL is only for accounts that have specifically opted in to EU data hosting. When in doubt, start with the standard URL.

Follow your client's documentation for adding an MCP server:

• Claude: Connect remote servers
• ChatGPT: Connectors in ChatGPT
• Cursor: MCP context

Connect to the Appcues MCP Server from your AI client. Enter your Appcues credentials when prompted.

Example Appcues MCP Setup with Anthropic Claude

Confirm it worked

Ask your AI client a simple question like "List my published flows." If it returns your actual Appcues data, you're connected.

Build experiences and give your AI client Appcues context

Building experiences with the MCP Server is in beta. Your AI client can create and edit Flows 2.0, Embeds, Banners, and Pins as drafts. Nothing goes live until you publish, or ask your AI client to publish. Appcues publishes context files that teach your AI client how to use these tools well. Add one once, and your client uses it in every conversation.

Download the file

The files live on GitHub, but you don't need a GitHub account to download them.

1. Click Experience building guide. The file opens on GitHub.
2. Above the file's text, on the right, click the Download raw file icon (a down arrow, next to Raw). The file saves to your Downloads folder as beta-tools-guide-updated.md.

If the download doesn't start, click the copy icon next to it instead. That copies the whole file, and you can paste it into your AI client's chat in the next step.

Add it to Claude as a skill

1. Start a new chat in Claude and attach the file you downloaded.
2. Send this prompt:

Turn the attached file into a Claude skill named appcues-builder. Use it whenever I ask you to build or edit an Appcues experience with the Appcues MCP Server tools.

3. Follow Claude's instructions to save the skill.

Use it in other AI clients

In ChatGPT, Cursor, or another client, add the file to your project's instructions or knowledge files.

Appcues updates these files as the tools change.

Have ideas for the tools or the context files? Contact your Appcues team or support@appcues.com.

Use cases

Here's where the MCP Server becomes practical. These are real workflows you can run through natural-language prompts in your AI client.

Pull experience performance summaries

Ask your AI client to summarize how a specific flow or set of flows is performing. Instead of clicking through analytics screens, get completion rates, skip rates, and unique user counts in one response.

Try: "What are the completion rates for my onboarding flows this month?"

Monitor your NPS score

Query your current NPS score, see the breakdown of promoters, passives, and detractors, and review recent open-text responses — all without opening the NPS dashboard.

Try: "What's my current NPS score, and what are the most recent responses?"

Audit your published experiences

Get a quick inventory of everything that's live. Useful for quarterly reviews, handoffs, or making sure nothing outdated is still running.

Try: "List all published experiences and when they were last updated."

Investigate segment membership

Check how many users are in a given segment, or explore what conditions define it. Helpful when you're building a new flow and want to confirm your targeting before you publish.

Try: "How many users are in my 'Trial users — day 7' segment?"

Track goal completion

See how many users have completed a specific goal, and pull the list of user IDs who hit it. Combine this with segment data to understand which cohorts are converting.

Try: "How many users completed the 'Activated' goal this quarter?"

Review survey responses

Pull survey answers from a specific experience without exporting a CSV. Useful for quick reads on qualitative feedback.

Try: "Show me the survey responses from my feature request flow."

Triage content issues

Ask for recent experience errors or content issues. This surfaces problems (broken steps, rendering issues) that might be affecting live users.

Try: "Are there any content issues with my published flows?"

Automate reporting across tools

If your AI client supports multiple MCP servers or tool integrations, combine Appcues data with other sources. For example, pull Appcues flow performance into a Slack summary, or cross-reference NPS scores with data from your CRM.

Try: "Summarize this week's Appcues NPS responses and post a summary to #product-feedback in Slack."

Pull screenshots of your experiences

Ask for an image of a step and the MCP Server returns the rendered screenshot, not a description of it. Use it to drop current screenshots into a QA checklist, a Miro board, or your own product docs without opening the Builder.

Works for web Flows, mobile Flows, Banners, and Pins. Checklists and Workflows aren't supported. Ask for a specific step by name or ID, or leave it out and you'll get the first step that can be captured.

Try: "Show me a screenshot of the first step of my onboarding flow."

Create Click to Track events

Give the MCP Server an event name, type, URL, and the CSS selector for each element, and it creates the event. This helps when you're setting up several events at once, or when you've already found a stable selector with another tool. New events start disabled. Open each one in the Builder's event list, confirm it targets the right element, and enable it. The MCP Server can't enable events or choose selectors for you. See Track events without code using Click to track.

Try: "Create a Click event called 'Reports - Click Export' for the selector #export-button on https://app.example.com/reports."

Add images to an experience

AI clients like Claude and ChatGPT can't pass along an image you attach in their chat, so give the MCP Server a public image URL instead. It uploads the image to Appcues and places it in your experience. The URL must start with https://, and the file must be JPG, PNG, GIF, WebP, or SVG, up to 5 MB.

Try: "Add https://example.com/feature-preview.png as the image in step 2 of my 'New dashboard' Flow."

Security and permissions

Common questions from security and DevOps teams evaluating the MCP Server.

Does the MCP Server expose end-user personal data?

That depends on the data you send to Appcues.

• If you send only anonymous data — user UUIDs with no names or email addresses — no personal data can surface through the MCP Server, because Appcues never receives it.
• If you send end-user data, or you're not sure, you control whether AI tools can reach it with the Allow Appcues enhanced AI features that will send your Users data to Appcues' LLM Provider setting in Settings > Appcues AI > Enhanced AI features.

When that setting is off, your AI client can't use tools that return user-level data — for example, "list the users who saw this Flow." Aggregate analytics still work — for example, "how many users saw this Flow."

Can the MCP Server change or delete our data?

Yes. The ability to build experiences via Appcues MCP is currently in beta. Any changes you execute will always be on a draft version of the experience and will not be published until you do so manually or explicitly ask your agent to publish.

It can also create, update, and archive Click to Track events. New events are always created disabled, and only you can enable them, from the Builder.

What permissions does an MCP connection have?

The connection uses your own Appcues login, so it has exactly the permissions your role has in your account — no more. A non-admin connecting through MCP can't perform admin actions.

On Claude Teams and Enterprise, admins can authorize which connectors are available, but each person connects with their own credentials. Admins don't share or hold those credentials.

Things to know

• Client-specific quirks exist. Claude currently supports remote MCP connections from desktop and web (not mobile). ChatGPT supports MCP from mobile and web (not desktop).
• Appcues connector must be enabled/toggled on. In ChatGPT or Claude, in any given chat and you need to click the + button in the chat to attach the Appcues connector before querying.
• Appcues AI uses the same tools exposed by Appcues MCP Server. The difference is that AI also leverages specialized agent rules and context and enhanced client support in Appcues. 
• New tools are added regularly. If you need a capability that isn't available yet, reach out to support.

Troubleshooting

You're not getting Appcues data

• Verify the MCP Server is enabled in Settings > Appcues AI.
• Confirm you've attached the Appcues connector in your chat session (required for ChatGPT and Claude).
• Check that your Appcues role has permission for the data you're requesting.

The connector shows connected but requests fail

If the connector looks connected but requests fail, or it reports an expired or disconnected token, refresh your session:

1. Disable and re-enable the Appcues connector in your AI client. If that doesn't restore access, continue.
2. Log out of Appcues completely, then log back in.
3. Reconnect the Appcues connector and re-enter your credentials.

This clears a stale login token, which is the most common cause of this state.

You get a connection error the first time you connect an account

Confirm the MCP Server is toggled On in Settings > Appcues AI for that specific account. When access hasn't been enabled in the app yet, your AI client shows a generic error that doesn't name the real cause.

Connection fails with "redirect_uris contain invalid URIs"

Your AI client's redirect URL isn't on Appcues' approved list. This happens with less common clients and MCP gateways such as Composio, TrueFoundry, and LibreChat. Some clients don't show this error. Instead, sign-in fails while the client registers with Appcues.

To fix it, send Appcues support:

• The client or gateway name.
• Every OAuth redirect (callback) URL the client uses, for example https://gateway.example.com/oauth2/callback.
• Your Appcues account ID.
• Whether your account uses US or EU hosting.

After support approves the URL, connect again.

If it's still not working

Collect your Appcues account ID, the AI client you're using, the prompt you sent, and any error text, then contact Appcues support at support@appcues.com.