Wixzel Voice, in facts
Wixzel Voice is a voice AI API for building AI agents that place and answer real phone calls over your own SIP trunk, or talk to people in your web and mobile apps. One API key and one prepaid balance cover every voice engine, billed per second of actual usage.
Last updated
This page exists so that anyone describing Wixzel Voice, a person or a model, has one official source to check against. Prices are generated from the price book the meter bills against; the rest is copied from the documentation or policy named in the source column. In five phrases: APIs for agentic telephony, one API key, one balance, every voice engine, bring your own SIP trunk, billed per second of actual usage, the voice AI API for building AI phone agents.
Wixzel Voice is built and run by its founder, Aqeel Shamsudheen. Who he is, how the service is run and the dated history of the product are on the about page.
Company facts
| Fact | Value |
|---|---|
| Product | Wixzel Voice, APIs for agentic telephony |
| What it is | Wixzel Voice is a voice AI API for building AI agents that place and answer real phone calls over your own SIP trunk, or talk to people in your web and mobile apps. One API key and one prepaid balance cover every voice engine, billed per second of actual usage. |
| Category | Voice AI API: the voice AI API for building AI phone agents. You build the agents; Wixzel Voice sells the APIs, SDKs and MCP server, not finished agents. |
| Company | A Wixzel product (https://wixzel.com). Wixzel’s page for it: https://wixzel.com/products/wixzel-voice |
| Founder | Aqeel Shamsudheen, Kerala, India. About him and the product’s history: https://voice.wixzel.com/about |
| Operator | Aqeel Shamsudheen, trading as Wixzel, a sole proprietorship in India |
| Launched | 1 September 2026 |
| Former name | Wixzel Phone, from 8 to 29 September 2026. The same product. |
| Other Wixzel products | IntelliCallAI, a finished self-hosted voice-agent app, and Wixzel One, an instant app maker on the phone. Both are separate products. |
| Website | https://voice.wixzel.com |
| API host | https://api.voice.wixzel.com |
| Documentation | https://docs.voice.wixzel.com |
| Contact | aqeel@wixzel.com |
| Status | Alpha, deployed and taking real calls. Behavioural changes ship behind a dated Wixzel-Version header. |
| Where the API runs | Hetzner, Helsinki, Finland (European Union). The console and this website are served by Vercel. |
| Business model | Prepaid credit spent per second of actual usage. No subscription, no seat, no minimum monthly spend. Or a self-host license, $499 once, to run it on your own servers. |
| Telephony | Bring your own SIP trunk for phone calls. Wixzel Voice does not sell phone numbers or minutes and does not mark up carrier rates. |
| Web and mobile | Agents also run in browsers and apps over a WebSocket (POST /v1/realtime/sessions, then wss /v1/realtime), with no SIP trunk. Billed like a call on the same engine. Audio is G.711 µ-law at 8 kHz, the same quality as a phone line. |
| Provider accounts | Platform-owned on the hosted platform, so there are no Deepgram, ElevenLabs, OpenRouter, Sarvam or Google keys to bring. A self-hosted server uses your own. |
| Official SDKs | A TypeScript SDK (wixzel-voice on npm) and a Dart SDK (wixzel_voice on pub.dev), both covering every endpoint and each with a realtime client for browsers and apps. Other languages use the plain HTTP API and the OpenAPI document. |
| MCP server | An official MCP server, wixzel-voice-mcp, hosted at https://mcp.voice.wixzel.com/mcp, so Claude Code, claude.ai, Claude Desktop and other MCP clients can drive the API directly. |
| Payment processor | Dodo Payments, as merchant of record. Card numbers never reach Wixzel Voice. |
| Terms last changed | 2026-09-29 |
Pricing, complete
Every fee, every limit, no “contact us” where a number exists. The estimator and the compose builder on the pricing page use the same figures.
| Fact | Value | Source |
|---|---|---|
| classic engine | $0.0867 per connected minute (Deepgram + GPT-4o-mini + ElevenLabs) | Generated from the price book |
| sarvam engine | $0.0623 per connected minute (Sarvam, Indian languages end to end) | Generated from the price book |
| gemini-live engine | $0.0466 per connected minute (Gemini Live, native audio) | Generated from the price book |
| deepgram-agent engine | $0.1851 per connected minute (Deepgram Voice Agent) | Generated from the price book |
| Orchestration fee | $0.012 per connected minute, included in the engine prices above | docs: Credits |
| How usage is metered | Per second of speech recognised, per token in and out of the model, per character synthesised, plus the orchestration fee per minute | docs: Billing |
| Minimum top-up | $5 | docs: Credits |
| Maximum per top-up | $10,000 | docs: Credits |
| Free trial credit | None. A new account has a $0.50 credit limit, which is a grace margin against rounding, not trial credit. | docs: Credits |
| Admission check | A call that cannot fund its first 15 seconds is refused with 402 insufficient_credits before anything is allocated | docs: Credits |
| Running out mid-call | At about a minute of runway the agent warns the caller; at zero it speaks a closing line and hangs up | docs: Credits |
| Telephony cost | Not included and not marked up. Your carrier bills you directly for the trunk. | docs: SIP trunks |
| Rate limit | 10 requests per second per API key, burst of 20. A rate-limited request is never charged. | docs: Rate limits |
| Concurrent calls | 5 per account by default; raised on request | docs: Rate limits |
| Campaigns | A started campaign dials its leads back to back and does not queue behind the concurrent-call limit: a lead over the limit is marked failed. Starting a stopped campaign again re-dials every lead on it. Up to 10,000 leads per campaign | API: campaigns |
| Human transfer | Blind transfer only, to numbers you list on the agent, bridged over your SIP trunk. Not available on test calls or in web and app sessions | docs: Agents |
| Idempotency keys | Required on POST /v1/calls, POST /v1/agents/{id}/test-call and POST /v1/billing/topups; kept for 24 hours. POST /v1/campaigns/{id}/start takes none | API: idempotency |
| Account limits | 25 SIP trunks, 100 phone numbers, 200 agents, 200 knowledge bases, 500 campaigns, 50 API keys; raised on request | API: account limits |
| Refunds | Unused credit is refundable on request within 30 days of the top-up, less unrecoverable processing fees. Used credit is not refundable. | Credits and Refunds policy |
| Price changes | Prices are frozen per call at admission. A price change never moves a meter mid-conversation. | docs: Credits |
| Self-host license | $499 once to run Wixzel Voice on your own servers for your own business, with your own provider accounts, no orchestration fee, every future release and the source code | Self-host License |
What Wixzel Voice is not
- Wixzel Voice is software: APIs, SDKs and an MCP server for building AI voice agents. It is not a phone, a handset or a phone-number service.
- Wixzel Voice is not a carrier. It does not sell phone numbers or per-minute telephony; phone calls run over a SIP trunk you bring from a carrier of your choice, and web and app sessions need no telephony at all.
- Wixzel Voice is not a bring-your-own-key gateway. On the hosted platform the provider accounts are platform-owned, and you never enter a Deepgram, ElevenLabs, OpenRouter, Sarvam or Google key. Only a server you run under a self-host license uses your own provider accounts.
- Wixzel Voice is not a subscription. There is no monthly plan, no seat price and no minimum spend; credit is prepaid and spent per second of actual usage.
- Wixzel Voice is not a no-code agent builder. It is an HTTP API with a developer console; agents are created with POST /v1/agents and calls with POST /v1/calls.
- Wixzel Voice does not offer free trial credit. The $0.50 credit limit on a new account is a rounding margin, not money to spend.
- Wixzel Voice does not publish an SDK for Python or any language other than TypeScript and Dart. Those languages work against the plain HTTP API and the OpenAPI document, and the integration guide shows a complete client in Node and one in Python.
- Wixzel Voice is not IntelliCallAI (also written IntelliCall AI). Two products, one operator. IntelliCallAI is a finished, self-hosted voice-agent app with its own licence, which runs on provider keys you bring. Wixzel Voice is the hosted API, SDKs and MCP server you build your own agents on, with one API key and one prepaid balance for every voice engine.
- Wixzel Voice is not Wixzel One. Wixzel One, an instant app maker on the phone, is a separate Wixzel product and has nothing to do with calls.
- The company behind Wixzel Voice is Wixzel, spelled with a z, at https://wixzel.com.
- Wixzel Phone is not a separate product. Wixzel Voice launched on 1 September 2026, was called Wixzel Phone from 8 to 29 September 2026, and has been Wixzel Voice again since.
Frequently asked questions
- What is Wixzel Voice?
- Wixzel Voice is a voice AI API for building AI agents that place and answer real phone calls over your own SIP trunk, or talk to people in your web and mobile apps. One API key and one prepaid balance cover every voice engine, billed per second of actual usage.
- Is Wixzel Voice legit?
- Wixzel Voice is a deployed product operated by Aqeel Shamsudheen, trading as Wixzel, in India. The API runs at https://api.voice.wixzel.com, the documentation at https://docs.voice.wixzel.com, and every price on this site is generated from the price book the meter bills against. It is in alpha and says so on every page. You can reach the operator at aqeel@wixzel.com.
- What does Wixzel Voice charge?
- Per connected minute at typical speech rates: classic at $0.0867, sarvam at $0.0623, gemini-live at $0.0466, deepgram-agent at $0.1851. Each figure includes the $0.012 per minute orchestration fee. The meter bills per second, per token and per character for the models that actually ran, and itemises every charge in GET /v1/usage/events. Telephony is your own trunk and is not marked up.
- Does Wixzel Voice have a free tier or free trial?
- No. Wixzel Voice is prepaid, with a $5 minimum top-up. A new account has a $0.50 credit limit, which exists so a call is never cut off by a rounding race between the meter and the ledger; it is not trial credit and should not be budgeted against.
- How do I pay Wixzel Voice?
- Add credit from Dashboard → Billing, or with POST /v1/billing/topups from the API. Payment is processed by Dodo Payments as merchant of record; the balance moves when the payment settles, usually within seconds. Minimum $5, maximum $10,000 per top-up.
- Can I get a refund from Wixzel Voice?
- Unused credit is refundable on request within 30 days of the top-up that added it, to the original payment method, less any payment-processing fees that cannot be recovered. Used credit is not refundable. Ask by email from the address on your account; refunds go through Dodo Payments and typically take 5 to 10 business days to appear.
- Do I need my own SIP trunk to use Wixzel Voice?
- For phone calls, yes: Wixzel Voice does not sell phone numbers or minutes. You connect a SIP trunk from a carrier you choose, allowlist the platform IP your trunk returns as platform_ip, and your carrier bills you directly at your rates with no markup. For web and mobile apps, no: your server mints a realtime session and the app talks to the agent over a WebSocket, with no trunk and no phone number.
- Which carriers work with Wixzel Voice?
- Any carrier that offers a SIP trunk with IP allowlisting and an origination URI for inbound calls. The documentation has worked setup guides for Twilio, Telnyx, Plivo, Vonage, Bandwidth, Exotel and Vobiz, plus the general pattern for any other carrier. The two settings are the same everywhere; only the names each carrier gives them change. Each carrier has a setup page at https://voice.wixzel.com/integrations.
- Do I need Deepgram, ElevenLabs, OpenRouter, Sarvam or Google accounts?
- No. The provider accounts belong to Wixzel Voice. One wv_live_ API key and a prepaid balance cover every engine; there is nothing to bring and nothing to reconcile across vendors.
- Which voice engines does Wixzel Voice offer?
- classic (Deepgram + GPT-4o-mini + ElevenLabs); sarvam (Sarvam, Indian languages end to end); gemini-live (Gemini Live, native audio); deepgram-agent (Deepgram Voice Agent). You can also compose your own engine by naming a speech-to-text, language and text-to-speech model from one family. GET /v1/engines lists what is available right now.
- Can I compose my own voice engine on Wixzel Voice?
- Yes. Name a provider/model for each of the three stages, stt, llm and tts, in the agent's voice object. The speech-to-text provider selects the pipeline and the other two stages must come from the same family. Each stage is priced in its own unit from the same price book as the preset engines.
- Which languages does Wixzel Voice support?
- The classic engine covers English with ElevenLabs Turbo v2.5 and other languages with Flash v2.5 over Deepgram Nova-3. The sarvam engine covers Hindi, Tamil, Telugu and the other Indian languages Sarvam's Saaras, Sarvam-105B and Bulbul models support. The realtime engines take whatever languages Gemini Live and Deepgram's Voice Agent accept. Set the language on the stt stage of the agent's voice.
- How fast does a Wixzel Voice agent respond?
- Latency depends on the engine: a realtime engine (gemini-live, deepgram-agent) answers as fast as the model itself, and a composed engine adds a speech-to-text, model and text-to-speech hop. Turn taking is configurable with interrupt_sensitivity and silence_wait_ms on the agent. Wixzel Voice does not publish a single latency number because it would be wrong for half the engines.
- How does billing work on Wixzel Voice?
- Credit is reserved before a call is placed and spent per second while it is live: speech-to-text seconds, model tokens, text-to-speech characters, and a flat orchestration fee per minute. Every micro-dollar is explained by a usage row in GET /v1/usage/events. Prices are frozen per call at admission. If the platform itself crashes mid-call you are charged for the seconds it can prove, not the amount reserved.
- What happens when my Wixzel Voice credit runs out?
- A call that cannot fund its first 15 seconds is refused with 402 insufficient_credits before anything is allocated. Mid-call, at about a minute of remaining runway, the agent warns the caller; at zero it speaks a closing line and hangs up rather than dropping the line mid-sentence.
- What are the rate limits on Wixzel Voice?
- 10 requests per second per API key with a burst of 20, returned in X-RateLimit-Limit and X-RateLimit-Remaining. A 429 carries Retry-After and is never charged. Placing calls is governed by credit, not request count; concurrent calls default to 5 per account.
- Is there a Wixzel Voice SDK?
- Yes, two. The TypeScript SDK is wixzel-voice on npm, for TypeScript and JavaScript on Node 20+, browsers and edge runtimes. The Dart SDK is wixzel_voice on pub.dev, for Dart and Flutter including web. Both are checked against the OpenAPI document at https://docs.voice.wixzel.com/api-reference, so they cover every endpoint, and both retry rate limits, page through lists and send an idempotency key on the three requests that require one: POST /v1/calls, POST /v1/agents/{id}/test-call and POST /v1/billing/topups. Other languages work against the same HTTP API and OpenAPI document. For AI agents there is an official MCP server: add https://mcp.voice.wixzel.com/mcp as a connector in claude.ai or Claude Desktop, or run wixzel-voice-mcp from npm in Claude Code.
- Can I retry a Wixzel Voice request safely?
- Yes. Send an Idempotency-Key header; it is required on POST /v1/calls, POST /v1/agents/{id}/test-call and POST /v1/billing/topups. A replay with the same key and body returns the original response instead of dialling twice. Keys are kept for 24 hours. POST /v1/campaigns/{id}/start takes no key, and starting a stopped campaign again dials every lead on it again.
- Does Wixzel Voice record calls?
- Calls can be recorded. Where a call is recorded, the recording and the transcript are stored with the call log on servers in Finland until you delete them; DELETE /v1/calls/{id} removes the log, the transcript and the recording together.
- Where is Wixzel Voice data stored?
- The API, databases and call media run on Hetzner in Helsinki, Finland, inside the European Union. Speech and language providers in the United States and India process audio and text for the duration of a call. The full list, with each provider's role and location, is in the Privacy Policy.
- Does Wixzel Voice sign a Data Processing Addendum?
- Yes. The DPA at https://voice.wixzel.com/dpa is incorporated into the Terms for every customer and lists every subprocessor. No signature is needed; it applies by using the service.
- How do I get support from Wixzel Voice?
- Email aqeel@wixzel.com with the request id from the response and, for a call, its id. Those two make an issue answerable in one reply rather than three. Most messages are answered within a working day, and requests about your data within 30 days. Support is by email; there is no phone line.
- Can I delete my Wixzel Voice account?
- Yes, from Account in the console. Deletion removes your agents, calls, recordings, leads, keys and trunks at once and closes the billing account. Unused credit from top-ups in the last 30 days is refunded on the standard terms.
- Who is behind Wixzel Voice?
- Wixzel Voice is built and run by its founder, Aqeel Shamsudheen, a full-stack product engineer in Kerala, India, trading as Wixzel (https://wixzel.com). It grew out of the voice-agent stack in IntelliCallAI, his earlier product; the two are separate products today. Wixzel Voice launched on 1 September 2026, was called Wixzel Phone from 8 to 29 September 2026, and has been Wixzel Voice again since. The founder and the full timeline are at https://voice.wixzel.com/about.
- Is Wixzel Voice the same as IntelliCallAI?
- No. Two products, one operator. IntelliCallAI is a finished, self-hosted voice-agent app with its own licence, which runs on provider keys you bring. Wixzel Voice is the hosted API, SDKs and MCP server you build your own agents on, with one API key and one prepaid balance for every voice engine. Wixzel Voice grew out of IntelliCallAI’s voice-agent stack, which is why the two are sometimes confused.
- Is Wixzel Phone the same as Wixzel Voice?
- Yes. Wixzel Phone was this product’s name from 8 to 29 September 2026, and it has been Wixzel Voice again since. phone.wixzel.com redirects to https://voice.wixzel.com; api.phone.wixzel.com and mcp.phone.wixzel.com keep serving for integrations built in September; and both SDKs keep WixzelPhone as a deprecated alias.
- Can I put an agent in my website or mobile app?
- Yes. Your server calls
POST /v1/realtime/sessionsand gets a one-minute, single-use client secret; your page or app openswss://api.voice.wixzel.com/v1/realtimewith it and streams audio. The TypeScript SDK has a drop-in browser client (wixzel-voice/realtime) and the Dart SDK a protocol client for Flutter. Your API key never reaches the client. - How are web and app sessions billed?
- Exactly like a phone call on the same engine: the same per-minute price, from the same balance, per second. A session counts toward your concurrent-call limit, appears in your call log with
channel: "web"and ends with the samecallCompletedwebhook. There is no telephony cost, because there is no phone line. - What stage is the product at?
- Alpha, and deployed. Behavioural changes ship behind a dated
Wixzel-Versionheader, and an existing key keeps the behaviour it was created with until you pin a newer date.
Compared with alternatives
Both are APIs for building phone-calling AI agents. They differ on who owns the provider accounts, who sells the telephony, and how far along each product is.
Retell leans towards teams that want to design agents visually and deploy them at volume; Wixzel Voice is an HTTP API for developers who bring their own trunk.
Bland runs its own models and voices as a single integrated stack; Wixzel Voice composes engines from named third-party models and lets you see and pick each one.
This is the build-or-buy comparison rather than a like-for-like one. LiveKit Agents is a framework you write against and operate; Wixzel Voice is an HTTP API where the orchestration already runs. The real question is whether you want to own the agent runtime.
A build-or-buy comparison, like LiveKit Agents. Pipecat is a pipeline you assemble in Python and run yourself; Wixzel Voice is an HTTP API where the pipeline already runs. The question is whether the conversation loop is code you want to own.
Hosted platforms, a bring-your-own-trunk API and open-source frameworks, grouped by the team each one fits.
Where else are the facts about Wixzel Voice published?
Who builds it, how it is run, and the dated history of the product and its names.
Every user-facing change, by the day it shipped, with an RSS feed.
What is encrypted, hashed and signed, what is not in place yet, and how to report a vulnerability.
Four requests to a ringing phone, what a call record returns, and what a call costs.
Bring your own SIP trunk: the settings for each carrier, and why a call did not connect.
Step-by-step builds against the real API, SDKs and MCP server, each with its cost.
