Integrations

MCP Integration

Connect ProdExam to AI coding assistants using the Model Context Protocol (MCP). Manage tests, analyze responses, and generate insights using natural language.

Secure OAuth 2.0 Authentication

Industry Standard

ProdExam uses OAuth 2.0 for MCP authentication. When you add the server URL, your AI assistant will open a browser window for you to sign in securely. No tokens to copy or manage!

Setup Guide

1. Register the ProdExam MCP Server

Run this in your terminal (project scope is committable to your repo):

claude mcp add --transport http --scope project prodexam https://testflow-fr0h.onrender.com/api/mcp

Use --scope user if you want it available across every project instead.

Or add it manually to .mcp.json at your project root:

{
  "mcpServers": {
    "testflow": {
      "type": "http",
      "url": "https://testflow-fr0h.onrender.com/api/mcp"
    }
  }
}

2. Start Claude Code

When Claude Code loads, it will open your browser to sign in to ProdExam. Approve the connection, and you're all set!

3. Verify Connection

Ask Claude: "List my ProdExam projects"

1. Open Cursor Settings

Press Cmd/Ctrl + Shift + P and pick View: Open MCP Settings (or Cursor Settings: Tools & MCPs). Click Add Custom MCP.

2. Add MCP Server Configuration

Add to ~/.cursor/mcp.json:

{
  "mcpServers": {
    "testflow": {
      "url": "https://testflow-fr0h.onrender.com/api/mcp",
      "transport": "http"
    }
  }
}

3. Restart and Authorize

Restart Cursor. When prompted, authorize ProdExam in your browser. Then ask: "What ProdExam tools are available?"

1. Install an MCP Extension

Install an MCP-compatible extension like "Continue" or "Roo Code" from the VS Code marketplace.

2. Configure MCP Server

Add to your extension's MCP configuration (varies by extension):

{
  "testflow": {
    "url": "https://testflow-fr0h.onrender.com/api/mcp"
  }
}

3. Reload and Authorize

Reload the window. When prompted, sign in to ProdExam in your browser. Then verify by asking about your ProdExam projects.

ProdExam uses OAuth 2.0 with PKCE for secure authentication. Any MCP-compatible client that supports OAuth can connect:

Server Details

MCP URL: https://testflow-fr0h.onrender.com/api/mcp

Transport: HTTP (JSON-RPC 2.0)

Protocol Version: 2024-11-05

Authentication: OAuth 2.0 with PKCE

Scopes: prodexam:read, prodexam:write

OAuth Discovery Endpoints

Authorization Server:
https://testflow-fr0h.onrender.com/.well-known/oauth-authorization-server

Protected Resource:
https://testflow-fr0h.onrender.com/.well-known/oauth-protected-resource

Available MCP Methods

initialize - Initialize the MCP connection

tools/list - List all available tools

tools/call - Execute a tool

user/info - Get authenticated user info

Available Tools

The MCP server provides 26 tools organized by category. Ask your AI assistant to use any of these naturally.

Projects

Tool	Description	Example Prompt
`projects_list`	List all accessible projects	"List my ProdExam projects"
`projects_get`	Get project details including API key	"Show me the API key for my checkout project"
`projects_create`	Create a new project	"Create a project called 'Mobile App'"
`projects_update`	Update project details	"Update the description of my checkout project"
`projects_delete`	Delete a project	"Delete the old test project"

Tests

Tool	Description	Example Prompt
`tests_list`	List tests with filtering by project/status	"Show all active tests"
`tests_get`	Get test details with stats and questions	"Get details for my checkout feedback test"
`tests_create`	Create a new test	"Create a test called 'Onboarding Feedback'"
`tests_update`	Update test (publish/pause/archive)	"Publish my draft test" / "Pause the checkout test"
`tests_delete`	Delete a test	"Delete the old test"
`tests_stats`	Get test statistics	"What's the completion rate for this test?"
`tests_add_step`	Add a question to a test	"Add an NPS question to my test"
`tests_update_step`	Update a test step/question	"Change the first question text"
`tests_delete_step`	Delete a test step	"Remove the last question"
`tests_share_link`	Create a public report URL for a test (requires card on file)	"Generate a share link for the checkout test results"
`tests_list_share_links`	List existing share links for a test	"Show the public links for this test"
`tests_delete_share_link`	Revoke a share link by ID	"Revoke the old share link"

Responses & Feedback

Tool	Description	Example Prompt
`responses_list`	List responses with sentiment/date filtering	"Show me all negative responses from the last week"
`responses_search`	Full-text search responses	"Find responses mentioning 'slow' or 'loading'"
`responses_by_question`	Group responses by question	"Show me responses organized by question"
`responses_trends`	Time-based trend analysis	"How has our NPS changed over the past month?"

Analytics & Insights

Tool	Description	Example Prompt
`analytics_dashboard`	High-level overview of all projects	"Give me an overview of my ProdExam data"
`analytics_sentiment`	Sentiment breakdown with examples	"What's the sentiment breakdown for this test?"
`analytics_completion_funnel`	Drop-off analysis by step	"Where are users dropping off in my test?"

Account

Tool	Description	Example Prompt
`account_info`	Get account and subscription details	"What plan am I on?"
`account_usage`	Get current usage statistics	"How many responses have I used this month?"

AI-Powered Analysis

Your AI assistant can perform advanced analysis by combining data from multiple tools. Just ask in natural language:

Feature	Example Prompt
Sentiment Analysis	"What's the overall mood of our users?"
Executive Summary	"Give me the TL;DR for my boss"
Theme Extraction	"What are people talking about?"
Priority Issues	"Which feedback needs urgent attention?"
Feature Requests	"What features are users asking for?"
Bug Detection	"Find any bug reports in our feedback"
Quote Extraction	"Give me 5 great quotes for our next demo"
Recommendations	"What should we fix first?"

These features use your AI assistant's reasoning abilities combined with your data - no special setup required.

Draft via MCP, fine-tune in the builder

The MCP is best for drafting — describe what you want and your AI creates a test with triggers and questions in seconds. When it's time to polish the details, every tests_create and tests_add_step response includes a builder_url that drops you straight into the visual builder for that test.

What you can fine-tune in the builder

• Triggers — re-capture by pointing at a real element on your app (no more guessing selectors)

• Question types — switch between multiple-choice, rating, NPS, free-text, yes/no, etc.

• Timeline — drag to reorder steps, set delays between them, add pauses

• Preview — run the test as a tester would (dry-run) without recording responses

• Publish — flip the test from draft to active once it's right

Typical flow

1. Ask your AI: "Create a checkout-friction test for my Pricing page"
2. It calls tests_create and tests_add_step for you
3. The response surfaces builder_url — click it to open the test in the dashboard
4. Adjust triggers and question types visually, preview in dry-run, then publish

Troubleshooting

"Tool not found" error

Make sure the MCP server URL is correct: https://testflow-fr0h.onrender.com/api/mcp. Restart your AI assistant after configuration changes.

Authentication error

Your OAuth session may have expired. Try removing and re-adding the MCP server, or visit Settings → MCP Access to check your connection status.

Browser doesn't open for authorization

Make sure your browser allows popups from your AI assistant. Check if the MCP server URL is correct in your configuration.

No data returned

Check that your test has responses. New tests start empty.

Connection timeout

The MCP server may be warming up. Try again in a few seconds.

MCP Connection Settings

How to Use AI Insights

Exporting Your Data