BYO Mode Explained: Bring Your Own API Keys to Voice AI

Published: January 19, 2026 Updated: January 19, 2026 Reading time: 10 minutes

What if you could just pay your AI providers directly?

No middleman markup on your OpenAI bill. No mystery fees on ElevenLabs usage. No platform taking a cut of every Deepgram minute. Just you, your API keys, and the actual provider pricing you negotiated.

That is BYO mode. And if you care about controlling costs at scale, it might be the most important feature you evaluate when choosing a voice AI platform.

Most voice AI platforms force you into their billing ecosystem. They provision API keys on your behalf, route your traffic through their accounts, and add a 15-40% markup on every LLM call, every TTS character, and every STT minute. For small deployments, this might seem like a convenience fee. At scale, it becomes your largest line item.

Burki's BYO (Bring Your Own) mode flips this model. You add your own API keys. You pay providers directly at your negotiated rates. You pay Burki only for the platform orchestration. Nothing else.

Let me show you exactly how it works and why it matters.

What Is Voice AI BYO Mode?

BYO mode is a billing configuration where you supply your own API credentials for AI providers instead of using the platform's managed accounts.

In standard "managed mode," a voice AI platform handles everything:

• They have accounts with OpenAI, ElevenLabs, Deepgram, etc.
• Your usage goes through their accounts
• They bill you for usage plus their markup
• You never see the underlying provider invoices

In BYO mode:

• You create your own accounts with AI providers
• You generate API keys from those accounts
• You add those keys to the voice AI platform
• Your usage goes directly to your provider accounts
• You pay providers directly at your rates
• You pay the platform only for orchestration

The difference is not just accounting. It is control.

How BYO Mode Works on Burki

Setting up BYO mode takes about 5 minutes per provider. Here is the process.

Step 1: Create Provider Accounts

You need accounts with the providers you want to use: an LLM provider (OpenAI, Anthropic, Google AI, Azure OpenAI, Groq, or xAI), a TTS provider (ElevenLabs, Deepgram, Cartesia, OpenAI, or Azure Speech), and an STT provider (Deepgram or Azure Speech).

Step 2: Generate API Keys

In each provider's dashboard, generate API keys. For Azure, you will also need the region identifier.

Step 3: Add Keys to Burki

Navigate to Organization Settings > Provider Settings. Select each provider, paste your API key, and save. Keys are encrypted at rest using AES-256 encryption.

Step 4: Configure Assistants

When creating an assistant, select your LLM provider/model, TTS provider/voice, and STT provider/model. The assistant uses your credentials for all API calls, and usage appears in your provider dashboards.

Per-Assistant BYO Keys

Here is something most platforms cannot do: Burki supports different API keys per assistant.

Why does this matter?

Client isolation: If you are an agency running voice AI for multiple clients, each client can have their own API keys. Usage is isolated. Billing is separated. No commingling of accounts.

Cost allocation: Different departments can use different keys for precise cost tracking.

Quota management: Spread usage across multiple provider accounts to avoid rate limits.

Testing vs. production: Use separate keys for development and production workloads.

You configure this in the assistant settings. Override the organization defaults with assistant-specific keys where needed.

Cost Comparison: Managed vs. BYO Mode

Let me show you the actual numbers. These are not hypotheticals. This is what voice AI costs look like.

Managed Mode Costs

In managed mode, Burki handles provider billing with a 15% markup on pass-through costs. This covers account management, support overhead, and consolidated billing convenience.

Here is a breakdown for a typical 3-minute call using GPT-4o-mini, Deepgram Nova 2 for STT, and ElevenLabs Turbo for TTS:

BYO Mode Costs

In BYO mode, you pay providers directly and Burki charges only the $0.03/min platform fee:

Savings at Scale

The per-call difference might seem small. Let me scale it up.

At 500,000 calls per month, BYO mode saves $66,000 annually. That is real money.

The Real Savings: Negotiated Rates

The numbers above assume you are paying standard provider rates. But here is where BYO mode really shines: you can negotiate.

At scale, providers offer volume discounts:

• OpenAI offers committed use discounts starting at certain spend levels
• ElevenLabs has enterprise pricing with significant per-character savings
• Deepgram offers volume pricing for high-usage customers

These discounts are tied to your account. In managed mode, you cannot access them because the usage goes through the platform's account. In BYO mode, your negotiated rates apply to every call.

One customer running 2 million minutes per month negotiated:

• 30% discount on Deepgram STT
• 25% discount on ElevenLabs TTS
• 40% discount on Azure OpenAI via an enterprise agreement

Their effective savings versus managed mode? Over 45% on provider costs. That translated to over $200,000 annually.

Setup Guide: Cost-Optimized BYO Stack

For the best balance of quality and cost, we recommend:

• LLM: GPT-4o-mini (OpenAI) - $0.15/$0.60 per million tokens
• STT: Deepgram Nova 2 - $0.0043/minute
• TTS: Deepgram Aura - $0.015 per 1,000 characters

This stack delivers sub-second latency at minimal cost. Setup takes 10 minutes:

1. Get your OpenAI API key from your OpenAI Platform dashboard
2. Get your Deepgram API key from console.deepgram.com
3. Add both keys in Burki's Organization Settings > Provider Settings
4. Create an assistant with OpenAI GPT-4o-mini, Deepgram Aura TTS, and Deepgram Nova 2 STT
5. Make a test call and verify usage appears in your provider dashboards

Mixing BYO with Managed: Hybrid Mode

You do not have to go all-in on BYO. Burki supports hybrid configurations where some providers use BYO keys and others use managed billing.

Common hybrid scenarios:

• Volume discounts on one provider: Use BYO for Deepgram (where you have a negotiated rate) and managed mode for ElevenLabs
• Testing new providers: Use managed mode to try Cartesia while keeping BYO for your primary stack
• Gradual migration: Start with STT on BYO, then move TTS and LLM over time

Each service category is billed independently:

In your assistant settings, you configure each provider independently. If BYO keys exist for a provider, those are used. If not, managed billing applies.

When to Use BYO vs. Managed Mode

Both modes have their place. Here is a framework for deciding.

Use Managed Mode When:

You are just getting started. Managed mode means zero setup friction. Create an assistant and start calling. No provider accounts required.

Volume is low. Under 5,000 calls per month, the markup is minimal in absolute dollars. Convenience may outweigh savings.

You want consolidated billing. One invoice from Burki versus multiple invoices from providers. Some finance teams prefer this simplicity.

You do not have provider accounts yet. Managed mode lets you test providers before committing to accounts.

Use BYO Mode When:

Volume exceeds 10,000 calls per month. The savings become significant enough to justify setup time.

You have negotiated provider rates. Those discounts only apply to your direct accounts.

You need precise cost allocation. BYO enables per-department, per-client, or per-project cost tracking.

You are an agency or platform. Client isolation requires separate keys.

You want maximum control. BYO means you own the provider relationship. No dependency on platform account status.

The Scale Inflection Point

In our experience, the inflection point is around 15,000-20,000 calls per month. Below that, managed mode is often fine. Above that, BYO starts generating meaningful savings that compound over time.

But if you already have provider accounts with credits or discounts, use BYO from day one. There is no reason to pay markup on resources you already own.

Frequently Asked Questions

Do I need BYO keys to use Burki?

No. Managed mode works out of the box with no API keys required. BYO is optional for users who want to control provider costs.

Can I switch from managed to BYO later?

Yes. Add your BYO keys at any time and configure assistants to use them when ready.

What happens if my BYO API key runs out of credits?

Calls will fail when the provider rejects requests. Burki does not fall back to managed mode automatically since that would bypass your cost controls.

Are my API keys secure?

Yes. Keys are encrypted at rest using AES-256 encryption and never logged in plaintext.

Does BYO affect call quality or latency?

No. The call processing pipeline is identical regardless of billing mode.

What about telephony?

Yes. You can bring your own Twilio, Telnyx, or Vonage credentials, or your own SIP trunk.

What providers support BYO on Burki?

All integrated providers support BYO:

• LLM: OpenAI, Anthropic, Google AI, Azure OpenAI, Groq, xAI
• TTS: ElevenLabs, Deepgram, Cartesia, OpenAI, Azure Speech, Inworld, Resemble AI, Kokoro
• STT: Deepgram (Nova 2, Nova 3, Flux), Azure Speech
• Telephony: Twilio, Telnyx, Vonage, BYO SIP

The Developer's Take on Cost Control

I have talked to hundreds of developers building voice AI applications. The ones who scale successfully have one thing in common: they understand their unit economics from day one.

Voice AI is not like traditional SaaS where costs are fixed. Every call has variable costs across multiple providers. At 10 calls a day, you can ignore this. At 10,000 calls a day, it is your largest expense after payroll.

BYO mode is not just about saving money, though it does that. It is about owning your cost structure.

When you use managed billing, you are subject to someone else's pricing decisions. If the platform increases their markup, your costs go up. If they change providers, your unit economics change. You have optimized for their cost structure, not yours.

When you use BYO, your costs are yours. You negotiate with providers directly. You optimize for your specific usage patterns. You benefit from volume discounts you earn. Your cost model is stable and predictable because you control it.

For developers building production voice AI systems, this control is not optional. It is the difference between a sustainable business and one that cannot afford to scale.

Getting Started with BYO Mode

Ready to take control of your voice AI costs?

Step 1: Sign up for Burki if you do not have an account. You get 200 free minutes to test.

Step 2: Create accounts with your preferred providers (OpenAI, Deepgram, ElevenLabs, etc.)

Step 3: Generate API keys and add them to your Burki organization settings

Step 4: Create an assistant using your BYO configuration

Step 5: Make test calls and verify usage appears in your provider dashboards

The entire setup takes under 30 minutes. The savings start with your first call.

Conclusion

BYO mode is the feature that separates voice AI platforms built for developers from those built for demos.

Managed billing is fine for getting started. But when you are ready to scale, when costs matter, when you need control, BYO is the answer.

Burki is the only voice AI platform that offers:

• Full BYO support for every integrated provider
• Per-assistant key configuration for client isolation
• Hybrid mode for gradual migration
• $0.03/min platform fee with no minimums
• Zero markup on BYO usage

Your API keys. Your provider rates. Your cost control.

That is how voice AI billing should work.

Ready to bring your own keys? [Start your free trial](https://burki.dev/signup) and configure BYO mode in minutes.

Related reading:

• Voice AI Platform Pricing Comparison 2026
• How to Reduce Voice AI Costs by 40%
• Understanding Voice AI Unit Economics