Introducing the Ear3 SDK: Embed AI Voice Interviews in Any Product
Third Ear TeamAugust 12, 20265 min read
Your users will tell you things in a conversation they'd never type into a survey. Getting people to talk was never the problem. The problem is pairing the fluidity and goal-directedness of a human interviewer with the infrastructure to make it scale — and that is brutally hard to build.
But along the way, we kept hearing the same thing from teams: "We don't want another research platform. We want this inside our product." A survey tool that wants voice alongside forms. A startup that needs AI interviews for customer discovery. A product team that wants conversational workflows built around its own users and data.
So today we're releasing the Ear3 SDK: the same infrastructure that powers Third Ear, now yours to build with. Drop a complete interview widget into your app, go headless for a fully custom UI, or drive everything from a CLI.
The hard parts are already done: interview protocols that stay on-goal, natural turn-taking, session management, low-latency audio, transcripts and recordings. You skip straight to the part where your product talks with your users.
Add a voice interview to your product in two minutes
We are releasing two packages:
@ear3/voice-interviewer— a turnkey React widget with a complete interview flow and sensible defaults (plus a headless layer for a fully custom UI).
@ear3/voice-config-cli — a CLI to the Third Ear platform: create interviews, list sessions, fetch keys, run ops from a terminal or CI.
Using Claude Code:
Below is the simplest way to add a voice interview functionality to your React App. If you already have an existing React app, skip the first step. Just hand it to your AI coding agent.
Prompt: Create an empty Next.js project. Add an Ear3 voice interview to my React app: install the @ear3/voice-interviewer package, create an interview with the ear3 CLI, about how much people use AI at work, three questions, one of them a 1 to 10 rating, and embed the component. Follow the package README
You might need to create an account and sign-in to get the API keys, and Claude will prompt you as needed. These instructions should work on Codex or your favorite AI coding tool.
Once the interview is live, every answer shows up in your dashboard (https://app.ear3.ai) — and you can ask Ear3 questions about the responses right from Claude Code.
Prefer to type it yourself?
Requirements: Node.js 20.9+ (download) — create-next-app refuses to run on anything older. npm and npx come bundled with Node.js; there is nothing separate to install.
Verify with node -v, npm -v, npx -v.
2. Install the CLI, authenticate and create an interview
npx @ear3/voice-config-cli login
npx @ear3/voice-config-cli create --name "AI at work" --prompt "Ask how much people use AI at work, three questions, one of them a 1 to 10 rating"
The CLI prints your interview ID (cm…) and your publishable key (pk_test_… / pk_live_…) right in the console. Copy both - you'll need them in step 3. You can always find them again on https://app.ear3.ai.
3. Put your credentials into the environment file
Next.js reads env vars from .env.local in the project root. Create that file with any text editor and paste these two lines, substituting the values you got in step 2:
Open your dashboard (https://app.ear3.ai), pick the interview and go to Responses. Every session lands there with the full transcript, the recording and per-question answers, plus the aggregated view across all respondents.
Once a few people have talked to it, switch to Insights and ask a question in plain language - "How often do people actually use AI at work, and what stops them?" - and Ear3 answers across all responses, with confidence and verbatim quotes. The same works from the terminal:
npx @ear3/voice-config-cli insights cm… "What are the main reasons people don't use AI at work?"
…or straight from Claude Code: "What did respondents say about AI at work?" - the Ear3 plugin calls the same insight engine.
That's it. AI generates the questions, WebRTC handles the voice, and every conversation turns into data you can actually query.
Build with Third Ear
We're excited to see what you build on top of them.
Building a product? Start with the SDK and docs. You can have goal-directed voice interviews running in your app this afternoon: customer discovery, onboarding, churn conversations, screening calls.
Scaling something serious? We work directly with teams on white-labeling, custom branding, and high-volume deployments. Email us at partner@ear3.ai.
Just exploring? Join our Discord. Tell us what you're making, request features, or show off what you shipped. We amplify the best builds.
