# BYOK Voice Agent Cost Math: ₹1 vs ₹1.20 vs ₹1.50 Platform Fee (2026)

> When bringing your own AI keys beats an all-in voice agent rate: how the ₹1, ₹1.20 and ₹1.50 per-minute BYOK platform fees work, with worked break-even examples.

Source: https://edesy.in/blog/byok-voice-agent-platform-fee-cost-math-2026

**Bring your own keys (BYOK) is cheaper than an all-in voice agent rate when your own AI provider cost per minute is below the plan rate minus the platform fee.** On Edesy the BYOK fee is from ₹1/min with all your own keys, ₹1.20/min with some, and ₹1.50/min standard. Telephony is extra either way, so compare only the AI part.

This post is the arithmetic. If you want the "why BYOK at all" overview first, read [the cheapest way to run an AI voice agent](https://edesy.in/blog/cheapest-way-to-run-ai-voice-agent-byok-2026). Here we work out where the break-even sits for your own call mix.

> **Run the numbers on real calls:** [Try Edesy free](https://voice-agent.edesy.in/?utm_source=edesy.in&utm_medium=blog&utm_campaign=byok-voice-agent-platform-fee-cost-math-2026) — ₹50 free credit, no card needed.

## The two ways to pay

**All-in (managed keys).** You pay one per-minute rate and Edesy runs the speech-to-text, LLM and text-to-speech on its own accounts. Deepgram speech-to-text and Google or Deepgram Aura voices are included. Premium providers are add-ons per minute.

**BYOK.** You paste your own provider keys into the workspace. The providers bill you directly, and Edesy charges only a platform fee for the call orchestration, tools, post-call extraction, campaigns and analytics.

All prices below are ex-GST (18% GST is added) and exclude telephony.

| Option | Per minute (AI part) | Who bills the AI providers |
|---|---|---|
| Pay as you go | ₹6/min | Edesy (included) |
| Pro plan (₹1,499/month, 300 min included) | ₹5/min | Edesy (included) |
| Max plan (₹4,999/month, 1,100 min included) | ₹4.50/min | Edesy (included) |
| Ultra plan (₹14,999/month, 3,500 min included) | ₹4/min | Edesy (included) |
| BYOK, all your own keys | from ₹1/min + your provider bills | You, directly |
| BYOK, some of your own keys | ₹1.20/min + your provider bills | Split |
| BYOK, standard | ₹1.50/min + your provider bills | You, directly |

Credit packs give the same rates without a monthly plan: Bronze ₹500+ at ₹6/min, Silver ₹5,000+ at ₹5/min, Gold ₹25,000+ at ₹4.50/min, Platinum ₹1,00,000+ at ₹4/min, with the rate locked at purchase. Full details are on the [pricing page](https://edesy.in/ai-voice-agent/pricing).

## The break-even formula

For any plan rate `R` and BYOK fee `F`:

**BYOK is cheaper when your provider cost per minute < R − F.**

| Compared with | Fee ₹1.00 | Fee ₹1.20 | Fee ₹1.50 |
|---|---|---|---|
| PAYG ₹6/min | under ₹5.00 | under ₹4.80 | under ₹4.50 |
| Pro ₹5/min | under ₹4.00 | under ₹3.80 | under ₹3.50 |
| Max ₹4.50/min | under ₹3.50 | under ₹3.30 | under ₹3.00 |
| Ultra ₹4/min | under ₹3.00 | under ₹2.80 | under ₹2.50 |

Your "provider cost per minute" is the sum of speech-to-text, LLM tokens and text-to-speech for one minute of a typical call. Get it from your own provider dashboards after a test batch, not from list prices, because LLM tokens depend heavily on prompt length, knowledge-base lookups and how chatty your callers are.

## Premium voices change the picture

The all-in rate includes Deepgram speech-to-text and Google or Deepgram Aura voices. If you want a premium voice on managed keys, it is an add-on:

| Managed add-on | Per minute |
|---|---|
| ElevenLabs TTS | ₹3.00 |
| Cartesia TTS | ₹2.00 |
| OpenAI TTS | ₹1.50 |
| Azure TTS | ₹1.20 |
| Google STT / Azure STT / Whisper STT | ₹1.00 / ₹0.80 / ₹1.20 |
| Custom or cloned voice | ₹5/min |

So the real comparison for an ElevenLabs agent on Ultra is ₹4 + ₹3 = ₹7/min, not ₹4. The Max and Ultra plans take 20% and 30% off add-ons, which narrows the gap, but this is the most common case where BYOK pays for itself: a team that already has an ElevenLabs or Cartesia account with good rates.

## Worked examples

These use **example provider costs** to show the method. They are not quotes of any provider's current price; check each provider's price page and your own invoices.

**Example 1: high volume, included voices.** A support line runs 20,000 minutes a month on Ultra-level rates with the included Deepgram and Google voices. Assume your own provider stack would cost ₹3.20/min. With all your own keys: ₹1 + ₹3.20 = ₹4.20/min, against ₹4/min all-in. **Stay all-in.** The included providers are already priced well, and you avoid managing three provider accounts.

**Example 2: premium voice.** An outbound sales agent uses ElevenLabs. On Max, that is ₹4.50 + (₹3.00 less 20%) = ₹6.90/min. Assume your own keys cost ₹4.00/min for the full stack. BYOK: ₹1 + ₹4.00 = ₹5.00/min. **BYOK saves about ₹1.90 a minute**, or roughly ₹19,000 on 10,000 minutes.

**Example 3: partial keys.** You bring your own LLM key (you have committed spend with one LLM provider) but use Edesy's speech providers. The fee is ₹1.20/min, plus the LLM tokens on your bill. This is worth it only if those tokens are close to free for you, for example from existing credits.

## Things the formula does not show

- **Rate limits are yours.** On BYOK, your provider's rate limits apply. If you run 50 parallel campaign calls, make sure your LLM and TTS accounts allow that concurrency. For Gemini, the Google AI Studio key pool of up to 10 keys spreads the load.
- **Failed keys fail calls.** An expired card on a provider account stops calls. Watch the call history for provider errors after any billing change.
- **Monthly plans still matter.** BYOK changes the per-minute AI cost, not the plan features. If you need unlimited agents, API access or more numbers, check which plan you need on the [pricing page](https://edesy.in/ai-voice-agent/pricing).
- **Telephony is a separate line.** It is billed separately in both models, so do not let a cheap or expensive carrier skew the AI comparison.
- **Billing pulse.** Wallet usage is billed in 60-second pulses, so very short calls cost the same as a full minute. Model your average billed minutes, not average talk time.

## How to decide in a week

1. Sign up, use the ₹50 credit and build the agent from a [template](https://edesy.in/ai-voice-agent/templates).
2. Run a test batch of real calls on managed keys and note the average billed minutes per call.
3. Add your own keys on a copy of the agent and run the same batch.
4. Read each provider's usage for that batch and divide by the minutes. That is your provider cost per minute.
5. Put it into the break-even table above.

The [BYOK pricing feature page](https://edesy.in/ai-voice-agent/features/byok-pricing) covers which keys are supported, and [bring your own keys and phone](https://edesy.in/ai-voice-agent/features/byok-byop) covers combining BYOK with your own SIP trunk or carrier. For per-provider rates on managed keys, see [provider costs in the docs](https://edesy.in/docs/voice-agent/billing/provider-costs).

## The short version

If you use the included voices and run under a few thousand minutes a month, all-in is usually simpler and close in cost. If you use a premium voice, already pay for provider credits, or run high volume with negotiated provider rates, BYOK at ₹1 to ₹1.50 a minute is where the savings are. Measure, then switch.

## Frequently Asked Questions

### What is the BYOK platform fee on Edesy?

From ₹1/min when every provider in the call (speech-to-text, LLM and text-to-speech) uses your own keys, ₹1.20/min when only some of them do, and ₹1.50/min as the standard BYOK fee. Prices are ex-GST; 18% GST is added. Telephony is billed separately.

### Who bills me for the AI providers when I bring my own keys?

The providers do. Your OpenAI, Gemini, Deepgram, ElevenLabs or other usage goes on your own account with that provider, at your own rates and credits. Edesy only charges the platform fee per minute.

### When is BYOK cheaper than an all-in plan?

When your own provider cost per minute is lower than the all-in plan rate minus the platform fee. For example, against the ₹4/min Ultra rate with a ₹1 fee, BYOK wins if your providers cost you under ₹3 per minute.

### Is telephony included in either option?

No. Telephony (the phone number and carrier minutes) is billed separately in both cases, so it does not change the BYOK vs all-in comparison. Compare the AI part on its own.

### Which keys can I bring?

OpenAI, Gemini, Anthropic, Deepgram, ElevenLabs, Sarvam, Soniox, Cartesia, AssemblyAI and Azure. For Gemini you can also add a Google AI Studio key pool of up to 10 keys.

### Do I need BYOK to start?

No. Most teams start on pay as you go at ₹6/min with the ₹50 sign-up credit, measure their real minutes and call mix, and only switch to BYOK once the numbers say it is worth it.