Providers & integrations

Every provider.
No lock-in.

Compose the best LLM, transcription, voice, and telephony for every job. Bring your own keys, use shared platform accounts, or disable a provider, decided per tenant.

Get a live callBrowse providers
TENANT_OWNED

Bring your own keys

Each tenant supplies its own encrypted provider credentials and pays providers directly.

PLATFORM_SHARED

Shared accounts

Tenants use a platform-owned provider account, ideal for agencies onboarding clients fast.

DISABLED

Policy control

Turn any provider off for a tenant. Credentials resolve dynamically, per call, at runtime.

Large language models

6 providers

Swap reasoning models per agent or per node, deterministic where you need it.

OpenAI
GPT-4o and GPT-4.1: flagship reasoning & tool use.
Popular
Anthropic
Claude models for nuanced, safe conversations.
DeepSeek
Cost-efficient reasoning at scale.
Llama
Open-weight models, self-hostable.
Cohere
Enterprise-grade retrieval & generation.
Mistral
Fast, efficient European models.

Speech-to-text

3 providers

Real-time transcription tuned for telephony audio.

Deepgram
Ultra-low-latency streaming ASR.
Popular
Azure Speech
Broad language coverage, enterprise SLAs.
Whisper
Open-source, self-hostable transcription.

Text-to-speech

5 providers

Natural, low-latency voices, pick the right one per agent.

ElevenLabs
Premium, expressive human-like voices.
Popular
Cartesia
Fast streaming TTS built for real-time.
Deepgram Aura
Low-latency voices, single-vendor stack.
AWS Polly
Reliable, cost-effective at scale.
Azure TTS
Neural voices with wide language support.

Telephony

4 providers

Place and receive calls with your own numbers and trunks.

Twilio
Global programmable voice & SIP.
Popular
Plivo
Cost-effective global voice APIs.
Exotel
Leading coverage across India & APAC.
Vonage
Enterprise voice & messaging.
Runtime credentials

Your keys. Your control.
One engine, many tenants.

Voxvol injects each tenant's provider credentials into the agent config at call time, so a single engine replica serves many tenants, each with its own stack, without global secrets.

Encrypted, per-tenant credential storage
Resolved dynamically per call, no env-var sprawl
Mix providers freely across LLM, STT, TTS & telephony
# resolved at runtime, per tenant
agent:
  llm: { provider: "openai", model: "gpt-4o" }
  transcriber: { provider: "deepgram", mode: "nova-2" }
  synthesizer: { provider: "elevenlabs", voice: "ava" }
  telephony: { provider: "twilio", did: "+1555…" }
  policy: "TENANT_OWNED" # or PLATFORM_SHARED

Compose your stack. Ship your agent.

Bring your own keys, swap any provider in a click, no lock-in, ever.

Get a live callRead the docs