# BYOK and Key Pools

> Bring your own API keys for OpenAI, Gemini, Deepgram, ElevenLabs, Sarvam, Soniox and other providers, add a pool of up to 10 Google AI Studio keys with failover, and pay a lower platform fee from ₹1/min.

Source: https://edesy.in/docs/voice-agent/workspace/byok

Bring your own key (BYOK) means your agents use your own provider accounts for the LLM, speech-to-text and text-to-speech, and Edesy charges a lower platform fee: from ₹1/min with all your own keys, ₹1.20/min with some, against the standard ₹1.50/min. For Google AI Studio (Gemini) you can add a pool of up to 10 keys, so calls fail over to another key when one hits its quota.

## Supported providers

| Provider | Used for |
|----------|----------|
| OpenAI | LLM, TTS, OpenAI Realtime |
| Google Gemini (AI Studio) | LLM, Gemini Live, Gemini TTS |
| Anthropic | Stored for features that use it. Claude is not a live-call LLM on the default runtime |
| Deepgram | STT (Nova, Flux), Aura TTS |
| ElevenLabs | STT (Scribe), TTS, voice cloning |
| Sarvam | STT, TTS |
| Soniox | STT, TTS |
| Cartesia | TTS, voice cloning |
| AssemblyAI | STT |
| Azure | STT, TTS, Azure OpenAI |

## Platform fee with BYOK

| Your keys | Platform fee per minute |
|-----------|-------------------------|
| All your own keys (LLM, STT and TTS) | From ₹1.00 |
| Some of your own keys | ₹1.20 |
| Standard BYOK platform fee | ₹1.50 |

Without BYOK, standard pricing applies: from ₹4/min on plans and credit packs, ₹6/min pay as you go.

Your providers bill you directly for their usage. Telephony is billed separately. See [billing](https://edesy.in/docs/voice-agent/billing) for plans and credit packs.

## Add a key

1. Open **Settings → API Keys** (provider keys).
2. Pick the provider and paste the key.
3. Save. Keys are stored encrypted and shown masked after saving.
4. Place a test call and check it in call history. If a key is invalid or out of quota, fix it in the same settings page.

## Google AI Studio key pool

| Item | Value |
|------|-------|
| Keys per workspace | Up to 10 |
| Strategies | **Priority** (use keys in order) or **Least busy** (spread calls across keys) |
| Session cap per key | Pre-filled at 10 concurrent Live sessions, adjustable |
| Failover | A key that fails (for example quota exceeded) is skipped and cools down; the next key is tried |

Use a pool when you run many Gemini Live calls at once and a single key would hit its concurrent-session or quota limits.

## FAQs

**Do I need keys for every provider?**
No. Bring keys only for the providers you want; Edesy provides the rest at the partial BYOK fee.

**Can I see what I would pay with BYOK?**
Ask [Copilot](https://edesy.in/docs/voice-agent/agent-setup/copilot) to estimate your BYOK per-minute cost, or see [provider costs](https://edesy.in/docs/voice-agent/billing/provider-costs).

**Are BYOK keys the same as vault credentials?**
No. BYOK keys are for AI and speech providers. The [credential vault](https://edesy.in/docs/voice-agent/tools/credential-vault) is for the APIs your custom functions call.

**Can I use a pool for providers other than Google AI Studio?**
Not yet. Other providers take one key each.

## Related

- [Workspaces and roles](https://edesy.in/docs/voice-agent/workspace/roles)
- [LLM providers](https://edesy.in/docs/voice-agent/llm)
- [Provider costs](https://edesy.in/docs/voice-agent/billing/provider-costs)