Tell Peekinduck about your software
Peekinduck needs the basics before it can demo anything: who the product is for, where it lives, and the questions you want asked on every call.
- Software profile · name, one-liner, who it's for
- Knowledge base · docs, marketing pages, FAQs, pricing
- Qualification questions · what Peekinduck should ask every prospect
Give Peekinduck the keys to a sandbox account
A throwaway login is all Peekinduck needs. Add custom login steps if your product has SSO or a multi-step flow.
- Sandbox credentials only · never use production data
- Login instructions · natural language, e.g. "click SSO, then choose Google"
- Encrypted at rest · used only by the exploration agent
Peekinduck maps your product the way a user would
Peekinduck sends you a video of every workflow
Same flow you'd run with a new team member — approve, send feedback, or ask for more.
Drop in up to 20 recordings
Loom exports, screen captures, whatever you already have. Peekinduck will turn them into a stitched, on-demand demo.
- Per file · ≤ 2 GB · ≤ 30 minutes
- Any source · Loom, Zoom, OBS, QuickTime, …
Mark which clip plays for each step
Peekinduck is live on your booking page
Every prospect that books a call now meets a voice agent that knows your product, asks the right qualifying questions, and demos the exact workflow the prospect cares about.
Create the agent, feed it the knowledge
The widget is the in-product voice agent your end-users talk to. Start with the same things you'd give a new support hire: who you are, what you do, where the docs live.
- Name + behavior · how the agent should sound and act
- Knowledge base · PDFs, help docs, FAQs, changelogs
- Status · processing → completed before going live
Lock down where it can run, pick how it looks
The widget is gated by your public key + an allowed-origins list — only the domains you whitelist can mount it. Theme it to match your product.
- Allowed origins · explicit allowlist, CORS-enforced
- Public key · safe to ship to the browser
- Theme · color + shape that fits your UI
Paste one script tag into your app
That's the whole install. The widget mounts itself, talks to Peekinduck's backend with the public key, and renders a launcher in your chosen corner.
- One tag · no build step, no SDK install
- Async · zero impact on initial load
- Auto-updates · we ship widget changes for you
<script src="https://peekinduck.ai/in-app-widget.js" data-public-key="oa_pub_5f7a3c2e9b1d4a86_nova_prod" data-api-base="https://api.peekinduck.ai" async ></script>
Teach the agent your product's actions
Want the agent to actually do things, not just point at them? Register a tool for every action it should be able to call — Peekinduck picks which one to run from the user's voice.
- Defined in your code · keeps execution on your side
- JSON schema · same shape OpenAI/Anthropic tools use
- Confirmation · flag destructive actions to need user approval
// Run any action the agent decides to call Voqi.defineTool({ name: "create_sprint", description: "Create a new sprint on the current board.", parameters: { type: "object", properties: { name: { type: "string" }, startDate: { type: "string", format: "date" } }, required: ["name"] }, execute: async (args) => api.sprints.create(args), requiresConfirmation: false });
Two modes out of the box
Voice built into your product
Every signed-in user now has Peekinduck in their corner — guiding new features, executing the actions you trust it with, and reporting back to you on what users actually try to do.