We are live on Product Hunt today. Try the server and tell us what works and what breaks.

Open the launch page

Cloud MCP server · OAuth · No installation

neuroflash in Claude, Cursor and your agents.

Let your AI agent ask 1,000,000 real people before it writes. Query your real audience and create on-brand content without leaving your chat tool.

The whole setup: tell your agent to install the neuroflash MCP. Or paste one URL and sign in via OAuth. Works with any MCP client.

real survey profiles
1,000,000+
prediction parity
85 to 98%
tools on the server
88+

What it looks like in your chat window

Two questions, two tool calls, an answer grounded in survey data instead of a guess.

neuroflash connector Claude Desktop

What does Gen Z think of our new claim "Build it like you mean it"?

twins.query · Segment: gen_z_de · 240 profiles

7.4/10. Sounds confident and direct, but several feel it lacks a concrete benefit. Recommendation: add a result.

Rewrite it in our brand voice with that fix.

content.generate · Voice: neuroflash

"Build campaigns like you mean it, and know in advance that they land."

The Digital Twins behind those answers are built on 1,000,000+ real survey profiles and 80+ studies. Prediction parity against the human samples they replicate averages 85% across published research and reached 98% in an internal client validation study.

Connected in about a minute

No install, no local process, no API key to store. The server runs in the cloud and authenticates with your neuroflash account.

The shortcut, in any agent system: paste this to your agent

Install the neuroflash MCP server: https://app.neuroflash.com/api/mcp-server/v1/mcp

Or by hand, in three steps:

  1. Open your tool

    Claude Desktop, Cursor, Windsurf or another MCP tool.

  2. Add the connector

    Settings, connectors, paste server URL and sign in via OAuth.

  3. Just ask

    "What does Gen Z think of our slogan?" The Digital Twins answer in chat.

Server URL, for Claude Desktop and any client with a connector field

https://app.neuroflash.com/api/mcp-server/v1/mcp

Cursor, in .cursor/mcp.json

{
  "mcpServers": {
    "neuroflash": {
      "command": "npx",
      "args": [
        "mcp-remote",
        "https://app.neuroflash.com/api/mcp-server/v1/mcp"
      ]
    }
  }
}

88+ tools · 3 interaction modes · OAuth 2.0 with PKCE

Today on Product Hunt

The Digital Twins MCP server is live on Product Hunt today. Questions, bug reports and blunt feedback are all welcome in the comments.

Per Product Hunt policy this page asks for feedback and support, never for upvotes.

neuroflash MCP - Let your AI agent ask 1M real people before it writes | Product Hunt

Questions

What is an MCP server?

MCP, the Model Context Protocol, is an open standard that lets an AI client call an external service directly. The neuroflash MCP server exposes Digital Twins, brand voices and content generation as tools your client can use, so the research and the writing happen in the same window.

Which clients work?

Claude Desktop, Cursor, Windsurf and any other MCP client that speaks HTTP streaming. In Claude Desktop you paste the server URL as a custom connector. Cursor goes through the mcp-remote bridge, shown in the snippet above.

What can I ask?

Anything you would put to a research team about an audience: reactions to a claim, a concept or a subject line, differences between segments, or a rewrite in your brand voice.

In the default mode the server offers 88+ tools plus a planning and a search layer, so an agent can chain several steps without you steering each one.

What does it cost?

MCP access follows your neuroflash plan, and every call draws on the same workspace quota as the web app. Current plans and limits are listed on the Digital Twins MCP page.

Is my data safe?

Sign-in runs over OAuth 2.0 with PKCE, so your neuroflash password never enters the AI client and you can revoke access from your workspace at any time. Redirect targets are restricted to approved hosts and loopback addresses.

neuroflash GmbH is based in Hamburg, Germany. The full details are in the privacy statement.