Building on Cartesia’s voice models, or installing a ready-made website agent?

How do Cartesia and Kawl Now compare?

Cartesia sells voice models, which other voice platforms such as Deepgram also use, and hosts agents you configure yourself. Kawl Now is a finished agent for one job. We checked Cartesia’s official site and docs on 8 October 2026.

What is Cartesia?

Cartesia describes itself as an AI lab building real-time voice models and agents. Sonic, its text-to-speech model, streams speech quickly enough for live conversation, and its latest version, Sonic 3.6, speaks 44 languages. Ink, its speech-to-text model, transcribes as people talk and detects when they have finished a turn. Both are sold through an API and paid for with monthly credits.

Other voice platforms use Cartesia’s voices too. Deepgram’s Voice Agent API and Ultravox both list Cartesia among their text-to-speech options, so you may hear Sonic inside a product that isn’t Cartesia’s own.

What are Cartesia Managed Agents?

Managed Agents combine Ink, a language model you choose and Sonic into a hosted agent that handles turn-taking and interruptions. You set the instructions, welcome message, language, voice and model in Cartesia’s playground or through its API. Agents can use system tools such as ending a call, webhook tools that call your server and client tools that run in your app, and each change that affects behaviour is saved as a version you can restore.

For knowledge, you upload documents into folders, where they’re indexed automatically, and attach folders to an agent. The agent searches only its attached folders through a built-in knowledge_base tool, and attaching or removing a folder takes effect at once, even on calls in progress. By default, Cartesia also records the audio and transcript of every call for review afterwards, shows each call’s logs and offers evaluations that score completed calls.

How would you put a Cartesia agent on your website?

Cartesia connects people to an agent through a phone number, or through a WebSocket that a web, mobile or server app uses to stream audio. Its docs don’t describe a website widget. In a browser, your server creates a short-lived access token with the agent grant, because the API key must never reach client code, and your page captures microphone audio as PCM, streams it in short chunks and plays the agent’s audio back.

Cartesia lists your own user authentication as a prerequisite for issuing those tokens, so a public website needs its own rules for who may start a call. The call screen, microphone handling and playback are also yours to build.

With Kawl Now, you paste one script tag and list the sites that may start calls. Visitors type what they need, allow the microphone and talk, with live captions, mute and an end button.

What does each cost?

On 8 October 2026, Cartesia priced Managed Agent calls at $0.06 a minute, plus $0.014 a minute on a Cartesia phone number. Its docs say that rate doesn’t include LLM usage, which is priced per model, though its pricing page listed LLM usage as free for a limited time on agents created in its UI. Plans bundle model credits with prepaid agent dollars: the free plan includes 20,000 credits and $1 of agent usage a month, about 17 minutes of calls, and paid plans run from Pro at $5 a month to Scale at $299 a month, with custom Enterprise pricing.

Kawl Now gives each workspace its first 50 conversations free, shared by support, sales and tests, with calls of up to 10 minutes. Paid calls aren’t available yet, so no payment is collected.

When is Cartesia the better choice?

Cartesia fits teams building voice into their own products. Look at it first if you need:

• Text-to-speech or speech-to-text inside your own app.
• Voice agents you design, with your choice of language model and tools that call your systems.
• Phone numbers, outbound or batch calls, or SIP trunking.
• Conversations in languages other than English, or a cloned brand voice.
• Paperwork for a security review: Cartesia says it is SOC 2 Type II certified and lists HIPAA and GDPR for its agents.

When does Kawl Now make more sense?

Kawl Now suits a SaaS team that wants spoken support and sales answers on its website without building a token server, the browser connection and the call screen. It has no phone number for customers to call, no model choice and no custom tools, and it greets callers in English. It gives you:

• Separate support and sales agents, each with its own knowledge, greeting, voice and widget.
• Agents that search only sources your team has approved, with GitHub docs reviewed again after each sync.
• Call limits handled for you: 10 minutes per call, three at once and 12 new calls a minute per workspace.
• Transcripts in the dashboard, and a REST API and MCP server where adding a source never approves it.

Common questions