MCP Server

What Is This?

Imagine being able to ask your AI assistant questions like:

• "What are the biggest UX problems on my site this week?"
• "Show me sessions where users abandoned checkout"
• "Get the code fix for this bug and open a pull request"

The SupaStory MCP Server makes this possible. It connects AI assistants to your SupaStory data so you can chat about your sessions and insights naturally.

MCP stands for "Model Context Protocol"—it's just a way for AI apps like Claude to connect to external data. You don't need to understand how it works, just follow the setup below.

Where It Works

Because the MCP server runs remotely on SupaStory's servers, it works with any app that supports the MCP protocol — no local installation or Node.js required:

• Claude Desktop — Anthropic's desktop app
• Claude Code — Anthropic's CLI for Claude, in the terminal
• VS Code (with GitHub Copilot or Claude Code) — Right inside your editor
• Cursor — The AI-powered code editor
• ChatGPT — OpenAI's chat interface
• Other MCP-compatible apps — Any platform that supports MCP

Paid plan required — The MCP server is available to users on a paid SupaStory plan. If you're on the free tier, you'll see a login error when you try to connect. Upgrade your plan to get access.

What You'll Need

• A SupaStory account on a paid plan with some recorded sessions
• An AI app that supports MCP — Claude Desktop is the easiest to get started with

Setup

{
  "mcpServers": {
    "supastory": {
      "url": "https://app.supastory.com/api/mcp/server"
    }
  }
}

claude mcp add supastory --transport streamable-http https://app.supastory.com/api/mcp/server

{
  "mcpServers": {
    "supastory": {
      "url": "https://app.supastory.com/api/mcp/server"
    }
  }
}

{
  "mcpServers": {
    "supastory": {
      "url": "https://app.supastory.com/api/mcp/server"
    }
  }
}

{
  "mcp": {
    "servers": {
      "supastory": {
        "url": "https://app.supastory.com/api/mcp/server"
      }
    }
  }
}

https://app.supastory.com/api/mcp/server

What Can You Ask?

Once connected, try asking things like:

Find Problems

• "What are the most critical issues in my app?"
• "Show me all the errors users hit this week"
• "Which bugs are hurting conversions the most?"

Understand User Behavior

• "What's frustrating users on the checkout page?"
• "Show me sessions where users got stuck"
• "Why are users abandoning the signup form?"

Get Fixes

• "Get the code fix for this issue"
• "Show me all issues that have fixes available"
• "Help me understand what this fix does"

Create Pull Requests

• "Save that fix and create a pull request on GitHub"
• "Turn this code fix into a PR I can review"
• "Fix this bug and open a pull request for it"

Manage Issues

• "Mark this issue as acknowledged"
• "What issues are still open?"
• "Dismiss this insight—we already know about it"
• "Show me which issues have been fixed"

Project Overview

• "How many sessions were recorded today?"
• "Give me a summary of issues by severity"
• "What pages are causing the most problems?"

The Insights Dashboard

When you ask about insights or issues, your AI app may display an interactive dashboard right in the conversation. This works on Claude Desktop, VS Code, ChatGPT, and other supported platforms.

The dashboard gives you a visual overview that's easier to scan than plain text:

• Summary badges at the top showing counts by severity (critical, high, medium, low)
• Severity filters to focus on what matters most
• Clickable insight cards with key details like affected users and whether a fix is available
• Detailed drill-down — click any card to see the full insight, including session info and code context

Try asking something like "Show me all critical issues" or "Give me an overview of this week's insights" to see the dashboard in action.

Not all MCP apps support interactive displays. If yours doesn't, you'll still get the same information as formatted text—no configuration needed.

Fix Issues and Create Pull Requests

The MCP server doesn't just show you problems—it can help you fix them and ship the fix, all from a conversation.

The AI handles the details—writing the PR description, choosing the right branch, and linking it back to the insight. You just review and merge.

Track Issue Status

As you work through your issues, you can update their status right from the conversation. This keeps your team in sync without switching to the dashboard.

Just tell the AI what to do:

• "Mark this as acknowledged—I'll fix it this sprint"
• "This one's fixed, close it out"
• "Dismiss this, it's expected behavior"
• "What issues are still open in my project?"

Available Tools Reference

Behind the scenes, the MCP server gives your AI assistant access to these tools. You don't need to know their names—just ask for what you want in natural language. But if you're curious about what's available, here's the full list.

Example Chat

Here's what a conversation might look like:

You: What's breaking on my checkout page?

Claude: I'll check your SupaStory data for checkout issues.

I found 3 problems affecting checkout:

1. **Payment button doesn't respond** (critical)
   - The submit button looks frozen on first click
   - 156 users affected this week
   - Code fix available

2. **Confusing error message** (high)
   - Users see "Error: undefined" when card is invalid
   - Affecting 23% of attempts

3. **Price mismatch** (medium)
   - Discount codes show different totals on cart vs checkout

Want me to get the fix for any of these?

You: Yes, fix the payment button issue and create a PR.

Claude: I'll get the code fix and create a pull request.

The fix adds a proper click handler and loading state to the
payment button. Here's what it changes:

- `src/components/CheckoutButton.tsx` — Adds onClick handler
  and disabled state during submission

I've created a pull request:
**PR #142: Fix unresponsive payment button on checkout**
→ https://github.com/your-org/your-repo/pull/142

The PR links back to the insight so your team has full context.
Want me to look at the other issues?

Troubleshooting

Claude doesn't see SupaStory

1. Make sure you saved the config file correctly with the URL https://app.supastory.com/api/mcp/server
2. Quit your MCP client completely and reopen it
3. Check that your config uses the "url" key, not the old "command" / "args" format

"Not authenticated" or login error Follow the authentication link provided by your MCP client to log in to your SupaStory account. Make sure you're on a paid SupaStory plan — free accounts don't have access to the MCP server.

Security & Privacy

Your data stays safe. Here's how:

Authentication is handled securely — When you log in through your MCP client, your session is authenticated directly with SupaStory's servers. Tokens are managed server-side.

Tokens refresh automatically — You don't need to log in again unless you explicitly sign out. The server handles token refresh in the background.

Access is limited to your projects — The MCP server can only see projects that belong to your SupaStory account. It can't access anyone else's data.

Personal information is scrubbed — Before session data is sent to the AI app you're using, the MCP server strips out personally identifiable information (PII) such as names, email addresses, and other personal details. This means the AI assistant never sees raw personal information from your recorded sessions.

No conversation logging — The server doesn't log your conversations, session data, or insights.

No third-party sharing — Your data goes between your MCP client and SupaStory's servers. It's never shared with third parties.