Introducing the OriginalVoices CLI

Today we released a command-line tool that gives developers and AI agents terminal access to our network of 25,000+ Digital Twins — AI representations of real people, each created, trained, and validated by the person it represents. Every response traces back to a real human, not a synthetic persona.

Installation and authentication take under a minute:

npm install -g @originalvoices/cli
ov auth login
ov ask "What matters most when choosing a new bank?" --audience "UK adults aged 25-40"

Why a CLI?

Shell scripts and autonomous agents all live in the terminal. Now they can talk to real people.

OriginalVoices already integrates with Claude, Cursor, and other AI tools through MCP. The CLI extends that to anywhere code runs — shell scripts and custom agent workflows. Developers get terminal-based access for quick studies, validation work, and piping real human perspectives into scripts.

What You Can Do

Ask anything, to anyone

Describe your audience in plain language. Ask open-ended or multiple-choice questions:

ov ask "What would make you pay for an AI fitness coach?" \
  --audience "Adults aged 25-35 who exercise 3+ times per week" \
  --sample-size high

Multiple-choice format:

ov ask "Which feature would matter most in an AI fitness coach?" \
  --type choices \
  --choices "Personalized plans,Form correction,Progress tracking,Nutrition advice" \
  --audience "Adults aged 25-35 who exercise 3+ times per week"

See what comes back

Every response comes from individual Digital Twins - real people with real perspectives:

{
  "question": "What would make you pay for an AI fitness coach?",
  "audience": "Adults aged 25-35 who exercise 3+ times per week",
  "twins_queried": 25,
  "responses": [
    {
      "answer": "Only if the AI could actually break down my specific combat technique and show me where I am leaking power or making mistakes in my footwork. It needs to work like a real coach who sees my sparring footage, points out what I am doing wrong, and gives me drills to fix it right away."
    },
    {
      "answer": "It would need to specifically personalize a training plan around my actual schedule, which is pretty chaotic with five kids and a 4am workout window. It would need to be affordable, show clear results for my fitness goals, and not require constant input from me once it is set up."
    },
    {
      "answer": "The main thing is replacing the need for constant manual tracking and planning by analyzing my specific data to provide actionable, daily workout adjustments. It needs to be more than just a generic template."
    },
    ...
  ]
}

This is what makes OriginalVoices different — you’re not getting a single AI-generated answer. You’re hearing from real people with different lives, contexts, and opinions.

Pipe output anywhere

All commands support --json for machine-readable output:

ov ask "What's your biggest frustration with online checkout?" \
  --audience "Frequent online shoppers" \
  --json | jq '.responses[].text'

Built for Agents

The CLI is designed for agentic workflows:

Structured output — --json support enables direct parsing without text scraping
Environment-based auth — OV_API_KEY allows headless authentication
Composable commands — Discrete, chainable operations
Natural language audiences — Plain English descriptions replace demographic codes

Getting Started

Requirements: Node.js 18+ and an OriginalVoices API key.

Install:
```
npm install -g @originalvoices/cli
```
Authenticate:
```
ov auth login
```
Grab your API key from platform.originalvoices.ai, or set the OV_API_KEY environment variable.

Ask your first question:

ov ask "What would make you switch from your current bank?" \
  --audience "US Adults, aged 30-45"

Full CLI documentation →