Using Kawl Now

What Kawl Now is, and what it isn’t.

Kawl Now is a browser voice AI agent for SaaS websites. A visitor clicks the widget, types what they need, allows the microphone and talks, and a support or sales agent answers from knowledge your team has approved. It has no phone number for customers to call and doesn’t answer phone calls, and it works beside your help centre, chat and onboarding tools rather than replacing them.

Read all 14 answers

What Kawl Now costs, and the limits that apply.

Each Kawl Now workspace gets its first 50 conversations free, shared by support, sales and your own dashboard tests, and the allowance is one-time rather than monthly. Every browser call stops at 10 minutes, and each workspace takes up to 3 calls at once. Paid calls aren’t available yet, so no card is required, nothing is charged and new calls stop when the 50 are used.

Read all 10 answers

Set up and test your website voice agent.

Join the waitlist and we set up Kawl Now with you: James opens your workspace, then you add and approve knowledge, configure the agent, list your website’s origin and paste one script tag just before the closing </body> tag. Customer calls start switched off, and you can test the agent from the dashboard before you add an origin or switch calls on. Each test that starts a voice session uses one of the 50 free conversations.

Read all 16 answers

What your agent knows, and how approval works.

Kawl Now agents answer from knowledge your team approves: FAQs and notes you write, plus selected .md, .mdx and .txt files from GitHub for the support agent. Adding a source never approves it, whether it comes from the dashboard, MCP or the REST API, and every GitHub re-sync needs approval again. Approval controls what the agent can use, but it doesn’t guarantee every answer, so test real questions and read the transcripts.

Read all 26 answers

What the support and sales agents can and can’t do.

Kawl Now gives each workspace two separate agents: a support agent that answers from approved FAQs, notes and GitHub docs, and a sales agent that answers from approved sales notes and never sees support sources. Each has its own name, greeting, instructions, voice, button and allowed origins. Neither can see a visitor’s account or the page they’re on, book meetings, update a CRM or call prospects.

Read all 25 answers

What visitors see and hear during a call.

Visitors talk to Kawl Now in their web browser, with nothing to install: they click the widget, type what they need, allow the microphone and talk. Before the call starts, the widget tells them they’re speaking with AI, that a transcript will be saved and that the call lasts up to 10 minutes. The agent greets callers in English in one of two voices, Marin or Cedar, and there’s no language setting.

Read all 11 answers

What your team sees after each conversation.

Kawl Now saves each conversation with its topic, status, length, a summary line taken from the caller’s own words and a two-sided text transcript, and it records no call audio. Automatic insights group recurring questions, reported issues and themes from up to your latest 30 completed customer calls for each agent in the last 30 days, with exact quotes from the transcripts. Kawl Now doesn’t run CSAT surveys or calculate resolution rates, so measure deflection and satisfaction with your own help desk figures.

Read all 9 answers

Manage your voice agent with MCP and the REST API.

Kawl Now has a remote MCP server and a REST management API, which an OpenAPI 3.1 file describes. Both take a workspace token from the dashboard, so a coding agent can add and approve knowledge, change agent settings and allowed origins, switch calls on, and read conversations and automatic insights. There’s no SDK, npm package or webhook, and no ready-made help desk, CRM or calendar connector.

Read all 13 answers

How Kawl Now handles conversation data.

Kawl Now saves each conversation as a text transcript in a Postgres database on Supabase in London, with row-level security on every table, and records no call audio. While a caller talks to the agent, audio passes between their browser and OpenAI, which processes the voice conversation, and after calls an OpenAI model analyses what callers said to produce automatic insights. OpenAI’s processing region isn’t documented, and there are no security certifications, retention settings or automatic deletion yet, so transcripts are deleted on request.

Read all 11 answers

Choosing and comparing

Voice agent, chat, phone line or your own build?

Choose by where your customers are when they ask: a website voice agent such as Kawl Now answers them while they’re on your site, an AI receptionist answers a phone number, and a chatbot or live chat answers in text. Kawl Now has no phone number for customers to call, so keep a phone or chat route for questions that need a person. Building your own agent means running a voice model, a server, document search, transcripts and spending limits yourself.

Read all 14 answers

How Kawl Now compares with voice agent platforms.

Kawl Now is a ready-made voice agent for one job: spoken support and sales answers on a SaaS website, from knowledge your team approves, with no phone number for customers to call. Platforms and APIs such as Vapi, Retell AI, ElevenLabs Agents, Ultravox and OpenAI’s Realtime API let developers build their own agents, often for phone lines too, so these answers compare price, website widgets, phone calls and keeping answers current. Vendor details come from each vendor’s own published pages, and each answer links to its full comparison with its sources and review date.

Read all 34 answers

How Kawl Now compares with chat and support tools.

Kawl Now adds spoken answers to a SaaS website and works beside your chat, help desk and callback tools rather than replacing them. These answers compare it with tools such as Fin, Chatbase, CallPage, Drift, Tidio, Crisp, Ada and Botpress on voice support, phone calls, passing a caller to a person and price. Vendor details come from each vendor’s own published pages, and each answer links to its full comparison with its sources and review date.

Read all 28 answers

Terms explained

Voice and speech terms, explained.

A voice agent has to hear what a caller says, notice when they’ve finished, reply aloud and cope when they interrupt, and each step has a name: speech recognition, endpointing, text to speech and barge-in. A chained agent uses a separate component for each step, while a speech-to-speech model hears and speaks by itself. Kawl Now runs on OpenAI’s GPT-Live voice model over WebRTC, so it has no separate speech recognition, text-to-speech or turn-taking settings to tune.

Read all 25 answers

AI and language model terms, explained.

A voice agent’s replies come from a large language model, and grounding the model in documents you trust, often through retrieval-augmented generation, keeps those replies close to your facts. Grounding reduces hallucinations but doesn’t remove them, so answers that matter still need checking. Kawl Now uses OpenAI models, searches your approved knowledge by keyword rather than by meaning, and can’t take actions in your product.

Read all 25 answers

Customer support and channel terms, explained.

Support teams judge self-service with measures such as CSAT, first contact resolution, average handle time and deflection, and each needs one fixed definition before you compare results. Channels differ in who answers and where: a person in live chat, a bot in a chat window, an IVR or AI receptionist on a phone line, and a website voice agent such as Kawl Now in the visitor’s browser. Kawl Now doesn’t calculate CSAT, resolution or deflection rates, so measure those with your own tools and its saved transcripts.

Read all 20 answers

Developer and data terms, explained.

Website voice agents are installed with embed code and managed through APIs, so a few developer terms come up early: API keys and bearer tokens authenticate requests, CORS decides which web pages can read a response, and MCP lets AI applications call a service’s tools. Data terms such as personal data, data processor and data residency decide what to check before a vendor handles customer conversations. Kawl Now installs with one script tag, offers a REST API and a remote MCP server that take a bearer workspace token, and sends no webhooks.

Read all 29 answers