Sautikit vs Africa's Talking: cheaper voice API numbers and per-second billing for Kenya

Sautikit and Africa's Talking are both Nairobi-based and both serve Kenyan developers building voice products. Africa's Talking (AT) is a multi-channel platform (voice, SMS, USSD, airtime, IoT) with a longer track record and broader African market coverage. Sautikit is voice-only, Kenya-first, and designed for teams that want a JSON-native API, M-Pesa STK push top-up, and per-second billing. It also ships AI natively: managed AI agents that answer inbound calls on your number and broadcasts that run outbound AI campaigns, both first-class API objects rather than something you assemble. Neither platform is the right answer for every use case. This post walks through the concrete differences so you can pick with open eyes.

All third-party pricing and feature claims are based on publicly available information as of 2026-07-17. Verify current figures on each provider's pricing page before making a commitment.

Africa's Talking was founded in 2010 and is headquartered in Nairobi. It is a multi-channel communications platform covering voice, SMS, USSD, airtime disbursement, payments, and IoT across more than 20 African countries. Voice is one component of a wider product surface. AT has direct relationships with mobile network operators across Africa and has processed substantial volumes across all its channels over its 15-year history.

Sautikit is a programmable voice API live in Kenya, built by Helloduty. It is voice-only by design: the API surface, documentation, and billing model are all optimised for teams building voice-first products, without the overhead of a sprawling multi-channel platform. When you do need SMS, USSD, or airtime alongside voice, they live one step away in the Helloduty family. Helloduty is the multi-channel CX platform (SMS, WhatsApp, USSD, agent desk) that Sautikit plugs into. Wallet top-up is via M-Pesa STK push. Billing is in KES.

Both companies are Nairobi-based. Both target developers. The difference is surface area and focus.

Prices below are as of 2026-07-17, per each company's public pricing pages. AT figures per founder confirmation / AT public pricing, 2026-06-30.

Item	Africa's Talking	Sautikit
Outbound voice (per min)	KES 4.50	KES 3.00
Inbound voice (per min)	KES 1.00	Free (KES 0)
Number rental	KES 5,000 setup + KES 2,000/month + VAT, assigned after a request	KES 100/month (ex. VAT) / KES 116 incl. VAT, live in seconds
Billing increment	Per minute	Per second after connect
Currency	KES	KES
Wallet top-up	Bank transfer, M-Pesa Paybill, USSD	M-Pesa STK push
Minimum balance	Verify on AT	None stated
Recording storage	Verify on AT	1 GB free

A few observations from the table:

The number rental contrast is the headline difference: AT charges KES 5,000 upfront plus KES 2,000 + VAT per month to hold a number, assigned after a request. Sautikit is KES 116 (incl. VAT) per month, live in seconds, with no setup fee and no minimum commitment.
AT's outbound rate (KES 4.50/min) is higher than Sautikit's (KES 3.00/min). Inbound calls on Sautikit are free (KES 0); AT's inbound rate is KES 1.00/min. For inbound-heavy or mixed workloads, Sautikit's effective blended cost is materially lower.
AT's M-Pesa top-up option goes through a Paybill number; Sautikit uses STK push (the phone prompts your M-Pesa PIN directly). STK push is faster for developers who want to top up in 30 seconds without navigating the M-Pesa menu.
Sautikit's per-second billing after connect means a 35-second call costs just over half a minute, not a full minute. For large call volumes, this matters.

FX note: both platforms bill in KES, so there is no USD/KES exchange exposure for either. This is a meaningful difference from global providers like Twilio.

AT and Sautikit use different formats to describe what should happen on a call (play audio, collect digits, transfer to an agent, etc.).

Africa's Talking uses XML-based Voice Actions (also accessible via a drag-and-drop Voice Builder in the dashboard). A simple "say something, collect a digit" flow looks like:

<?xml version="1.0" encoding="UTF-8"?>
<Response>
  <Say>Press 1 for account balance, or 2 to speak with an agent.</Say>
  <GetDigits timeout="30" numDigits="1" callbackUrl="https://yourapp.example/digit" />
</Response>

Sautikit uses JSON natively. The equivalent flow:

{
  "actions": [
    {
      "say": {
        "text": "Press 1 for account balance, or 2 to speak with an agent.",
        "language": "en-KE"
      }
    },
    {
      "getDigits": {
        "numDigits": 1,
        "timeout": 8000,
        "action": "https://yourapp.example/digit"
      }
    }
  ]
}

Both approaches describe the same call graph. The practical difference is that JSON is easier to generate programmatically from code: you can build a voice flow from a database-driven menu without constructing XML strings. For teams generating flows dynamically (personalised IVR menus, language selection, etc.), JSON tends to reduce boilerplate.

Sautikit accepts both formats at runtime: the JSON DSL (parsed and validated by Sautikit) and raw XML (forwarded to the telephony engine byte-for-byte). Teams migrating from AT or Twilio can keep their existing voice XML; JSON is optional, for when you want Sautikit-side validation or typed SDKs.

Multi-channel from one bill. If you need SMS, USSD, airtime disbursement, and voice under a single provider, AT bundles all of that. Sautikit takes a different path: it keeps voice clean and focused, and the other channels live in the Helloduty family. Helloduty gives you SMS, WhatsApp, USSD, and an agent desk that pairs with Sautikit voice, so you stay in one billing family without a single monolithic platform to navigate.

Track record. AT has been running since 2010 and has processed high volumes across its channels. If your team weights vendor age heavily, that history is worth noting. Sautikit is newer and purpose-built for the Kenyan voice workloads teams ship today, with the Helloduty group behind it.

African markets beyond Kenya. AT covers 20+ countries with direct carrier relationships. Sautikit is Kenya-first by design; that focus is what keeps its pricing and support sharp for the Kenyan market. If your roadmap needs voice in Uganda, Ghana, or Tanzania alongside Kenya, talk to the Helloduty team about multi-market coverage rather than defaulting away.

Inbound and outbound rates. Sautikit's inbound calls are free (KES 0/min) and its outbound rate is KES 3.00/min, lower than AT's rates (KES 1.00/min inbound, KES 4.50/min outbound) as of 2026-07-17. For inbound-heavy or mixed workloads, Sautikit is the cheaper option.

USSD. AT bundles USSD directly. With Sautikit you keep voice focused and add USSD through Helloduty when your product needs that channel: same family, no monolith required.

Voice-only focus. If you are building a product that only needs programmable voice, Sautikit's narrower scope means there is less surface to learn, fewer abstractions, and a simpler mental model. The documentation is entirely about voice. The API is entirely about voice.

M-Pesa STK push top-up. Topping up a Sautikit wallet takes about 30 seconds: call POST /v1/topups with provider: "mpesa" and your phone number, your phone rings with an M-Pesa PIN prompt, done. No Paybill menu navigation required. For developers who want to automate wallet management or fund wallets quickly during development, STK push is faster.

JSON-native voice flows. Sautikit's voice actions are JSON objects. This is easier to generate from code, serialize from a database, or build with a typed SDK. XML construction in code is more error-prone.

Per-second billing after connect. A 35-second call costs you 35/60 of a minute, not a full minute. For high-volume outbound calling (OTP calls, fraud alerts, appointment reminders), this adds up to meaningful savings versus per-minute billing.

Recording included free. Up to 1 GB of recording storage is included; paid tiers start at KES 599/month for 10 GB. Verify AT's recording pricing before comparing total costs.

Outbound rate. Sautikit's outbound rate (KES 3.00/min) is lower than AT's outbound rate (KES 4.50/min, per founder confirmation / AT public pricing 2026-06-30). Inbound calls are free on Sautikit (KES 0/min), which means mixed inbound/outbound workloads cost significantly less than outbound-only pricing implies.

Broadcasts for outbound. AI agents for inbound. Sautikit ships both as first-class API objects: create an agent with POST /v1/agents (a name, a base prompt, a model from Sautikit's catalog, a voice), publish an immutable revision, and bind it to a number you already own:

{
  "inbound_agent": {
    "enabled": true,
    "agent_id": "9d2b1f53-8c0e-4f1d-9a6b-5d3a8c47e9f0",
    "mode": "after_hours",
    "schedule": {
      "timezone": "Africa/Nairobi",
      "hours": [{ "days": [1, 2, 3, 4, 5], "start": "08:00", "end": "17:30" }]
    },
    "handover_callback_url": "https://yourapp.example/voice"
  }
}

That is the whole deployment: PUT /v1/numbers/{id}/routing with the object above and the agent answers your line. mode: "always" covers every call; mode: "after_hours" (shown) keeps your team answering during business hours and puts the agent on duty at night and over the weekend. On the call itself, callers can interrupt the agent mid-sentence, recording is on by default, and per-turn transcripts come back from GET /v1/calls/{id}?transcript=1. When the agent decides a human should take over, Sautikit POSTs a CallHandover webhook carrying the agent's reason and a summary of the conversation so far, and you reply with ordinary JSON voice actions — dial your support line, enqueue, whatever fits. The AI briefs your team before a human picks up. And if every AI seat is busy or the agent is unpublished, the call falls through to your normal routing: callers never hear an error.

Outbound is the same product pointed the other way. POST /v1/agents/{id}/send places a one-off agent call, and POST /v1/broadcasts runs an entire campaign: a CSV of contacts, a prompt template with per-contact variables like {{name}}, calling-hour windows, retries on no-answer, busy, failed, or voicemail-detected dials, and a report.csv when it finishes.

Pricing follows the model the rest of this page describes: AI time is KES 4.00 per minute, metered per second, on top of the standard per-second call rates, drawn from the same prepaid M-Pesa wallet. No new billing relationship, no USD. The free tier includes 1 concurrent AI call; paid tiers run 5, 8, or 10. And because the number underneath goes live in seconds at KES 116/month incl. VAT, the distance from "no phone system" to "an AI answering a dedicated line" is one afternoon — not a KES 5,000 setup fee, a KES 2,000 + VAT monthly commitment, and a provisioning queue before you can begin.

Assembling the equivalent elsewhere means wiring speech recognition, a model, speech synthesis, and telephony together yourself, then operating the result. On Sautikit it is three API calls — or none: the hosted MCP server exposes create_agent, publish_agent, update_number_routing, and the broadcast tools, so an AI assistant can stand up the whole receptionist for you. Start with AI agents, broadcasts, or the step-by-step inbound agent guide.

Pick Sautikit if:

You want a number now. Claim one from the console or API and it is live in seconds at KES 116/month incl. VAT — no request queue, no setup fee, no provisioning wait. On AT you request a number, pay KES 5,000 upfront plus KES 2,000 + VAT per month, and wait for it to be assigned.
You already have AT voice XML. Sautikit runs AT-style XML byte-for-byte at runtime, so existing call flows move without a rewrite — and the JSON DSL is there when you want flows that are easy to generate programmatically.
You want SDKs built for generating flows from code. npm install @sautikit/node gives a typed client plus voice-action builders that compose IVR flows as plain objects — no XML string assembly — and @sautikit/webrtc puts calling directly in the browser. Everything else is clean JSON over HTTP from any language.
You bill clients of your own. Instant numbers with no setup fee leave room to resell: an agency or ISV can spin up a workspace and a dedicated number per client the same day, at a cost that fits inside a client invoice. A KES 5,000 setup fee per number makes that model hard to run anywhere else.
Per-second billing matters for your call profile. Outbound is KES 0.05 per second from the moment the call connects; inbound is free.
You top up the way Kenya pays. M-Pesa STK push funds the wallet from the console or API — no Paybill menu, and low-balance webhooks let you auto-recharge.
You want AI answering the phone. Managed AI agents answer inbound calls on your number — every call, or only outside business hours — and broadcasts run outbound AI campaigns, at KES 4.00/min metered per second from the same wallet. Prefer to run your own model? The Stream verb forks live call audio over WebSocket for a realtime agent (Gemini, OpenAI) you operate yourself.
You want recording included. Up to 1 GB of call recording storage at no charge.
Your AI tools can run it. Sautikit ships a hosted MCP server with 62 tools, so Claude or any MCP client can claim numbers, place calls, create and publish AI agents, rewire number routing, and launch broadcasts — with human confirmation on anything that spends money. Nothing comparable exists on Africa's Talking today.

Africa's Talking's genuine edge today is breadth: one vendor for SMS, USSD, and airtime across multiple African markets. If that is the deciding factor, pair Sautikit for voice with Helloduty for SMS, WhatsApp, USSD, and an agent desk — the same coverage, with a voice layer that provisions in seconds instead of a queue.

If you are moving a voice integration from Africa's Talking to Sautikit, the main changes are:

1. Keep your XML (adjust verb differences) or adopt the JSON DSL.

Sautikit accepts both at runtime, so you can keep your existing AT voice XML; just fix a couple of verb differences: Sautikit's <GetDigits> nests its prompt <Say> inside the element and uses maxDigits/action, where AT uses numDigits/callbackUrl.

AT XML (works on Sautikit after the verb tweaks):

<Response>
  <Say>Hello. Your balance is ready.</Say>
  <GetDigits maxDigits="1" timeout="30" action="https://yourapp.example/choice" />
</Response>

Or adopt the JSON DSL if you want Sautikit-side validation (return from your voice_callback_url webhook handler):

{
  "actions": [
    { "say": { "text": "Hello. Your balance is ready.", "language": "en-KE" } },
    {
      "getDigits": {
        "numDigits": 1,
        "timeout": 8000,
        "action": "https://yourapp.example/choice"
      }
    }
  ]
}

2. Replace the AT authorization header.

AT uses a username/API key pair in basic auth or a Bearer token depending on the endpoint. Sautikit uses a long-lived ES256 JWT API key:

# Africa's Talking
-H "apiKey: <AT_API_KEY>"
 
# Sautikit
-H "Authorization: Bearer <SAUTIKIT_API_KEY>"

3. Update callback URLs and webhook verification.

Sautikit signs webhooks with HMAC-SHA256 over body + "." + timestamp, delivered in the X-Sautikit-Signature: t=<ts>,v1=<hex> header. Update your webhook handler to verify this signature instead of AT's mechanism.

4. Update number provisioning.

AT's number provisioning flow differs from Sautikit's. In Sautikit: GET /v1/numbers/available to browse, POST /v1/numbers/{id}/claim to claim. Numbers are KES 100/month (ex. VAT) recurring.

5. Update the outbound call endpoint.

# Sautikit outbound call
curl -X POST https://api.sautikit.com/v1/calls \
  -H "Authorization: Bearer $SAUTIKIT_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "from": "+254712345678",
    "to": ["+254700000001"],
    "voice_callback_url": "https://yourapp.example/voice"
  }'

The to field is an array of E.164 strings. from must be a number owned by your workspace.

Create a Sautikit workspace and claim a phone number.
Top up over M-Pesa: KES billing, no card.
Point your first AT voice flow at Sautikit (keep the XML or use the JSON DSL) and place a test call.

Start with Sautikit → · See pricing → · Need SMS, WhatsApp & an agent desk? Helloduty →