The Conversational AI Platform built for developers
Launch enterprise-grade voice agents in minutes, not days. Build, test, and deploy using the CLI, Git, and CI/CD workflows you already trust.
RemyMaison Lumiere
Find a table, handle a special request, confirm the booking.
Calls are recorded to improve Dialog-RSN-1. They’re not used for training. See our Privacy Policy for more information.
Your stack. Your workflow. Your call.
The PolyAI developer platform is designed around the workflows your team already has, not the other way around. Integrate via API, or take full control with the ADK.
PolyAI solves ASR, TTS, barge-in handling, and sub-second latency. You write the business logic.
Dialog-RSN-1
The fastest, most accurate, safest and cost-efficient reasoning model for real-time dialog—live on customer calls today
- Fastest<300ms median model first token
- Most accurateAudio native reasoning
- SafestGrounded, tool-gated
- Cost-efficientRequest-based serving
Call Dialog-RSN-1 from your own stack
Our audio-native dialog model, on its own. Get a key in a couple of minutes, or let your AI coding assistant install the CLI and set everything up for you.
Get set up from your AI assistant
Works in Claude Code, Cursor, Codex and Copilot CLI. You need a GitHub account. Your key never passes through the assistant.
Show the prompt
Set me up with PolyAI’s Dialog-RSN-1 API. Install the PolyAI CLI, sign me in with GitHub, and save my API key to my environment. Never print the key or ask me to paste it into this chat.
We solve the hard problems. You own the business logic.
Ultra-low latency conversational AI, lifelike TTS, and barge-in handling, so your AI calling agents handle real human conversations without missing a beat.
Enterprise-grade voice quality
Lifelike, ultra-low latency, and handles real human interruptions. Powered by PolyAI's proprietary ASR (automatic speech recognition) and TTS (text to speech) engines.
Local-first development
Edit agent flows, functions, and voice config entirely in your local environment. No browser UI required. Works with any IDE or editor.
Git-backed workflows
Every agent resource is a versioned file. Branch, diff, review, and merge agent changes exactly like application code. Full history, rollbacks included.
Deep API integrations
Define custom functions that call your backend APIs in real time. Trigger Stripe refunds, look up live order status, authenticate users, all within a call.
Built-in testing
Write and run automated test suites for your agent from the CLI. Catch regressions before they reach production. Drop it into your existing CI pipeline.
CI/CD ready
The poly push command is first-class CI/CD compatible. Ship agent updates in the same release cycle as your application code.
“The PolyAI ADK unlocked a massive increase in our development velocity. By bringing our voice agents into standard, Git-backed local workflows, our engineering team can iterate, validate, and push updates across multiple projects with incredible speed.”
What to expect from our developer platform
- 45languages spoken
- 85%Typical CSAT score
- 99.99%Uptime
Need answers?
How do I get a Dialog-RSN-1 API key?
Sign up at studio.poly.ai, or copy the setup prompt into Claude Code, Cursor, Codex or Copilot CLI. The prompt installs the PolyAI CLI, signs you in with GitHub and saves your key for you.
What does the Dialog-RSN-1 API cost?
See the API tab on the pricing page. New accounts start with free credit.
What is the Agent Development Kit and how does it work?
The ADK is a developer-facing SDK and CLI tool that lets your team build Agent Studio-compatible voice AI agents from your native IDE. Rather than clicking through the Agent Studio UI to configure an agent, you use an LLM (eg Claude) to drive the underlying Agent Studio API, so you can describe what you want to build in natural language and have it construct the agent's data structures and configuration for you. The ADK connects to a GitHub repository, so your team gets full version control and a familiar code-first workflow.
Does the ADK replace Agent Studio, or do the two work together?
They work together seamlessly. The ADK operates at the API level: it creates and modifies agents in Agent Studio under the hood. You can build or change something in the ADK, push it, and immediately see the result in the Agent Studio UI. You're encouraged to jump between the two: the IDE is where you build quickly, and the Agent Studio UI is where you visualise flows, review branches, and do your testing and staging before going live.
Does the ADK work with any IDE?
The ADK is agnostic to any IDE. PolyAI provides a set of rules built from agent design best practices. These are now tool-agnostic, so they work with whatever AI agent or coding environment your team already uses.
Can I visualise the agent workflow I’ve built through the ADK?
Yes. Once you push changes from the ADK, you can open Agent Studio to see the full visual representation of your workflow: how each step connects, where branches lead, and how the flow is structured. You can also use Claude within your IDE to generate a visual or textual representation of the workflow before pushing it.
Can I push external data into the agent to personalise the conversation?
Yes. The ADK supports hooks within the agent workflow that allow you to push third-party data in at runtime. This data can then influence how the agent responds and what path through the workflow it takes, for example, passing in a caller's website visit history to tailor the opening of the conversation.
Have a question that isn’t covered here? Ask us on Discord.
Start building your first voice agent today
Get going right now with ADK and API offerings or build your own Agent with the Agent Studio platform