MCP Server Originalvoices

What ovlabs/mcp-server-originalvoices MCP server does

The ovlabs/mcp-server-originalvoices MCP server connects MCP-compatible clients to OriginalVoices AI twins for audience research. It lets an agent describe a target group and submit multiple questions in one request. The intended output is audience feedback that can add human context to research, creative development, and decision-making.

The repository identifies OriginalVoices as the service behind the AI twins and provides an MCP interface rather than a standalone audience database or analytics system. The available operation is centered on asking a selected audience questions.

How it works

An MCP client invokes the ask_twins tool with two parameters:

• audience: a string describing the target audience
• questions: an array of questions to ask that audience

For example, an agent could describe a group as UK fitness enthusiasts aged 18–30 and ask about priorities when buying running shoes and replacement frequency. The server passes the request to OriginalVoices AI twins and makes the response available to the calling agent.

The ovlabs/mcp-server-originalvoices MCP server does not require a separate audience identifier in the documented interface. The audience is specified directly as text for each request, which suits exploratory research prompts where the target segment changes between questions.

Setup and configuration

Install the published npm package with npm install @originalvoices/mcp-server, or configure a client to run it with npx. The documented Claude Desktop configuration starts the package as an MCP server and supplies credentials through the environment.

Set the required ORIGINALVOICES_API_KEY environment variable to an API key created on the OriginalVoices Platform. The key is required for the server to access the service. When using Claude Desktop, add an originalvoices entry to claude_desktop_config.json, set the command to npx, and pass the package as an argument. The API key belongs in that entry's env object.

The repository also documents running a built local copy with Node and dist/index.js. From source, install dependencies with npm and build the project before starting the generated file. The source configuration still requires ORIGINALVOICES_API_KEY.

Tools and capabilities

ask_twins is the documented tool. It supports:

• Defining an audience in natural-language text
• Sending one or more questions as a string array
• Requesting audience research from OriginalVoices AI twins

This interface can support early-stage audience exploration, reaction gathering for creative concepts, and questions about preferences or behaviors. The material does not document filtering by demographics beyond what is written in the audience description, result formats, batch limits, or additional tools.

Limitations and notes

The ovlabs/mcp-server-originalvoices MCP server requires an OriginalVoices API key, so access depends on the OriginalVoices Platform and its API availability. The provided documentation does not state pricing, request quotas, response latency, privacy terms, or how closely AI twin responses represent any particular individual.

The documented tool accepts an audience description and questions, but no other controls are specified. Treat returned responses as audience-research input rather than verified survey results unless the service documentation provides additional validation guidance. The repository is licensed under MIT.