Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.poly.ai/llms.txt

Use this file to discover all available pages before exploring further.

You’re in Agent Builder mode. This quickstart walks the self-serve path on the Agentic Dialog Platform: Agent Builder scaffolds the agent, you embed via Web Calling. Click the Free trial — exit pill in the top-right corner for the full enterprise quickstart.
The fastest path from sign-up to a working voice agent on your site. Agent Builder does most of the building: you describe what you want, and it scaffolds flows, knowledge, voice, and integrations for you.
1

Create your account (~1 min)

Go to studio.poly.ai and sign up with your email or Google account. No sales call to start.
PolyAI sign-up page with email form and Google SSO option
You land on the Agent Studio home page. Agent Builder asks for your company’s website. Paste a URL and it starts building a tailored agent immediately.
Agent Studio onboarding screen asking for your company website URL
The platform is self-serve from here. No billing, no metering, no contract.
2

Describe your agent (~3 min)

Tell Agent Builder what you want to build. A URL, a document, or a sentence describing the business is enough to get started.Example prompts:
  • Build a reservations agent for a restaurant. Take name, party size, date, and time, then confirm.
  • Read https://example.com and build an agent that answers questions about the company.
  • I run a clinic. Build an agent that handles appointment requests and FAQ about opening hours.
Agent Builder reads your website, asks a few clarifying questions, then proposes a full agent plan: flows, knowledge, voice, and integrations. Review the plan and click Approve.
Agent Builder researching a website and planning a new voice agent
Agent Builder asking interactive questions about the agent's capabilities
Agent plan card with Revise and Approve buttons
See Prompting Agent Builder for prompt patterns that work well.
3

Refine knowledge and behavior (~3 min)

Agent Builder will have populated Build > Knowledge > Managed Topics based on what you gave it. Open it and skim:
  • Topics: each is a question/answer pair grouped by intent. Edit any topic to tighten the answer or add sample utterances.
  • Sources: upload extra PDFs or paste URLs to expand the knowledge base. Agent Builder turns them into topics automatically.
Ask Agent Builder to make changes in plain language, “add a topic for refund policy”, “make the welcome message warmer”, “don’t quote prices”. It edits the agent for you.
Adding a knowledge base topic in Agent Studio
Changes stay as Drafts until you publish to Sandbox for testing.
4

Test in-browser (~2 min)

The fastest way to test. Uses your device’s microphone over WebRTC. No number to dial, no app to install.
  1. Click Test in the top-right corner to open the Agent Debugging side panel
  2. Choose Call agent under Voice
  3. Select Sandbox from the environment dropdown
  4. Begin speaking to your agent
Call agent side panel with environment set to Sandbox and a Start test call button
Every test conversation lands in Conversations. Read the transcript, check what topics matched, and ask Agent Builder to fix anything the agent got wrong.
5

Embed or share (~1 min)

Publish your agent and put it in front of users.
Embed a click-to-call voice widget on any website. WebRTC end-to-end, no number to dial, no app to install.
  1. Go to Configure > Web Calling in the sidebar
  2. Copy the embed snippet
  3. Paste it into your site’s HTML
See Web Calling for styling, behavior, and channel-mix options.
Promote through environments when you’re ready:
  1. Go to Deployments > Environments in the sidebar
  2. Click Promote to Live to make the current Sandbox build production-ready
Some projects use a simplified pipeline that promotes directly from Sandbox to Live. You can roll back to any previous version. See the deployment pipeline guide for details.

​
Next steps

Master Agent Builder

Prompt patterns and the plan-review workflow.

Embed Web Calling

Voice widget setup, channel-mix matrix, and styling.

Build flows

Multi-step workflows for bookings, forms, and structured tasks.

Build locally with the ADK

Pull your project as YAML and Python, edit anywhere, push back.
Create an account, build an agent, add knowledge, test it, and deploy. Five steps.

​
Prerequisites

  • A use case in mind (e.g., customer support, reservations, FAQ)
1

Create your account (~1 min)

Go to the sign-up page and create your PolyAI account.
  1. Click “Sign up with Google”
  2. Select your Google account or enter your credentials
  3. Click “Continue” to authorize PolyAI
Google SSO sign-up screen
Once signed in, you’ll land on the Agent Studio home page.
2

Create your agent (~2 min)

From the home page, click + Agent to start the agent creation wizard. You can create a blank agent or import an existing configuration.Configure the basics:
  • Agent name, internal identifier for your project
  • Response language, primary language for responses (see multilingual support for additional languages)
  • Voice, select from available text-to-speech (TTS) voices
  • Welcome greeting, first message users receive (can be customized later in agent settings)
Click Next to enter Agent Studio.
Agent creation wizard with name, language, voice, and greeting fields
You can also duplicate an existing agent by clicking the three-dot menu next to any agent on the home page.
3

Add knowledge (~5 min)

Navigate to Build > Knowledge > Managed Topics in the sidebar.Click Add topic and provide:
  • Topic name, what this topic covers (e.g., “Store hours”)
  • Sample questions, up to 20 ways users might ask (e.g., “When are you open?”)
  • Answer, the response your agent should give
Click Save to create the topic.
Adding a knowledge base topic in Agent Studio
Changes are saved as Drafts. Publish to Sandbox to test them. Learn more about environments and versions.Optional: Add more topics to expand your agent’s capabilities. You can also:
  • Upload PDFs or URLs to auto-generate topics
  • Connect external knowledge sources like Zendesk or Google Sheets using the Connected tab in Knowledge
  • Add actions to trigger handoffs, SMS, or other behaviors
See the full Managed Topics guide for details on how RAG (retrieval-augmented generation) powers topic matching.
4

Test your agent (~2 min)

The fastest way to test, uses your device’s microphone directly.
  1. Click the phone icon in the top-right corner
  2. Select Sandbox from the environment dropdown
  3. Begin speaking to your agent
In-browser call test panel
Testing tips:
  • Use specific keywords to trigger your agent’s topics
  • Test with different accents and speaking styles
  • For multilingual agents, switch languages mid-conversation to test detection
  • Review conversations in the Conversations dashboard after testing
5

Deploy to production (~1 min)

Promote your agent through the deployment pipeline:
  1. Go to Deployments > Environments in the sidebar
  2. Click Promote to Pre-release for user acceptance testing (if available in your project)
  3. Click Promote to Live to make your agent production-ready
Each environment can have its own phone number and configuration.
Deployment pipeline showing Sandbox to Live promotion
Some projects use a simplified pipeline that promotes directly from Sandbox to Live, skipping Pre-release. You can roll back to any previous version if issues arise, see the deployment pipeline guide for details.

​
Next steps

Add advanced features

Connect APIs and add dynamic behavior with Python functions

Build conversation flows

Multi-step workflows for bookings, forms, and structured tasks

Configure voice settings

TTS, voice selection, and audio settings

Monitor performance

Dashboards, conversation review, and metrics

​
How your agent works

Each conversation turn follows a pipeline: ASR (speech to text) → LLM (knowledge retrieval and response generation) → TTS (text to speech). See architecture for a detailed breakdown, or processing order for the step-by-step flow.
Last modified on May 18, 2026