# Alternatives to VoiceAgents

> Source: https://theagentsindex.com/voiceagent/alternatives (re-verified)

Best match first.

## Shortlist

### 1. ElevenLabs Agents

Hosted platform for building voice and chat agents, with telephony, SDKs and a choice of LLM.

**Where it improves on VoiceAgents:**
- [Done-for-you service: partial — Self-serve software you run, but enterprises can buy ElevenLabs Forward Deployed Engineers who scope, build and stabilise the deployment; no price is published.](https://elevenlabs.io/agents/forward-deployed-engineers)
- [Multi-agent: yes — An orchestrator agent routes a live conversation to specialist agents on natural-language conditions, and transfers can be nested; the transcript carries across.](https://elevenlabs.io/docs/eleven-agents/customization/tools/system-tools/agent-transfer)
- [Agent permissions: yes — Tool use is gated per MCP server and per tool: always ask, fine-grained (auto-approved, requires approval, disabled), or no approval at all.](https://elevenlabs.io/docs/eleven-agents/customization/tools/mcp)

**Trade-offs against VoiceAgents:**
- None found.

[View ElevenLabs Agents](https://theagentsindex.com/elevenlabs-conversational-ai)

### 2. Vapi

Developer platform for building, testing and deploying voice AI agents that make and receive phone calls.

**Where it improves on VoiceAgents:**
- [Done-for-you service: partial — You build and run the agent yourself; the Scale contract adds an embedded Vapi engineer for the go-live, not a done-for-you build.](https://vapi.ai/)
- [Multi-agent: yes — Squads route one conversation across several specialised assistants with context-preserving handoffs.](https://docs.vapi.ai/quickstart/introduction)
- [Open models: yes — Open-weight models reach the pipeline through Groq, Together AI, Mistral, DeepSeek and OpenRouter, or a self-hosted server.](https://vapi.ai/platform)

**Trade-offs against VoiceAgents:**
- [Opt out of training: partial — The privacy policy reserves call content for model training and names no toggle; the $1,000/month ZDR add-on is the only documented way to stop it being kept at all.](https://vapi.ai/privacy)

[View Vapi](https://theagentsindex.com/vapi)

### 3. Cartesia

Real-time speech models and a managed runtime for production phone and web voice agents.

**Where it improves on VoiceAgents:**
- [Open models: partial — The LLM layer routes through LiteLLM to 100+ providers, but the vendor names only Anthropic, OpenAI and Google, and its own speech models are API-only.](https://docs.cartesia.ai/line/sdk/agents)
- [On the command line: yes — A `cartesia` CLI creates, deploys and manages agents, and can chat with agent logic locally.](https://docs.cartesia.ai/line/cli)
- [Self-hosted: partial — On-premise or VPC model deployment is sales-gated; separately, agent reasoning code can run on your own servers with Cartesia connecting to your URL.](https://www.cartesia.ai/pricing)

**Trade-offs against VoiceAgents:**
- [Opt out of training: partial — Opt-out is by request form for selected categories of content; Enterprise plans can additionally enable Zero Data Retention for TTS and STT inference.](https://www.cartesia.ai/legal/privacy)

[View Cartesia](https://theagentsindex.com/cartesia)

### 4. Deepgram Voice Agent API

One WebSocket API that runs speech-to-text, LLM orchestration and text-to-speech for real-time voice agents.

**Where it improves on VoiceAgents:**
- [Multi-agent: partial — Handoff between specialised agents is a documented pattern with a reference implementation you run yourself: your orchestrator opens a fresh Voice Agent session per agent and summarises context between them. The API itself has no built-in agent-to-agent coordination.](https://developers.deepgram.com/docs/multi-agent-architecture)
- [Open models: yes — Open-weight LLMs are listed as supported models: gpt-oss-20b through Groq, which requires you to supply the endpoint, and an NVIDIA Nemotron model that Deepgram manages. Deepgram's own speech and voice weights are not distributed.](https://developers.deepgram.com/docs/voice-agent-llm-models)
- [On the command line: partial — The dg CLI covers transcription, synthesis, text analysis and account management; its own getting-started enumeration and command reference contain no voice-agent command, while the agentic-tools page claims the CLI reaches voice agents. Both statements are recorded in notes.](https://developers.deepgram.com/developer-tools/cli/getting-started)

**Trade-offs against VoiceAgents:**
- [Free tier: partial — A $200 signup credit with no credit card required, spendable on any endpoint including the agent, plus a hosted playground. It is a credit grant, not a free tier: once it is spent, every call is metered.](https://deepgram.com/pricing)

[View Deepgram Voice Agent API](https://theagentsindex.com/deepgram-voice-agent-api)

### 5. LiveKit Agents

Open-source Python and Node.js framework for voice AI agents, with LiveKit Cloud hosting, telephony and observability.

**Where it improves on VoiceAgents:**
- [Multi-agent: yes — Multiple agents in one session via handoffs, plus a supervisor pattern where one agent keeps control and delegates to scoped tasks.](https://docs.livekit.io/agents/logic/workflows.md)
- [Open models: yes — Open-weight models are the recommended default in LiveKit Inference (Gemma 4 31B, GPT OSS 120B, DeepSeek), and the Ollama plugin runs models locally.](https://docs.livekit.io/agents/models/llm/)
- [On the command line: yes — The lk CLI creates projects, deploys agents and manages LiveKit Cloud; agents can also be run and talked to locally in console mode.](https://docs.livekit.io/reference/developer-tools/livekit-cli.md)

**Trade-offs against VoiceAgents:**
- [Pooled budget: partial — No per-user balances: allowances pool inside a project. Sharing one plan across projects, and quotas managed at workspace level, are Enterprise only.](https://docs.livekit.io/deploy/admin/quotas-and-limits.md)

[View LiveKit Agents](https://theagentsindex.com/livekit-agents)

### 6. PolyAI

Enterprise platform for building voice and chat agents that resolve customer calls, priced per minute.

**Where it improves on VoiceAgents:**
- [Done-for-you service: partial — Self-serve software plus vendor-delivered work: PolyAI builds and maintains managed dashboards, configures managed-service integrations, and runs an enterprise build of roughly six weeks.](https://docs.poly.ai/analytics/dashboards/custom)
- [Open models: partial — The model catalogue enumerates only Raven, OpenAI and Amazon Bedrock (Claude, Nova) with no open-weight option, but you can attach your own LLM endpoint.](https://docs.poly.ai/behavior/models/byom)
- [On the command line: yes — The poly CLI pulls, branches, validates, pushes and chats against a project from the terminal.](https://docs.poly.ai/extend/adk)

**Trade-offs against VoiceAgents:**
- [Pooled budget: partial — Conversation volume draws on one account-level Subscription Volume of CX Tokens, but the Wren builder allowance is per person and not shared.](https://docs.poly.ai/wren/usage-and-limits)
- [Opt out of training: partial — Customer conversations reach Raven training only where the contract authorises it, and the privacy notice relies on consent where law requires; no self-serve opt-out control is documented.](https://docs.poly.ai/legal/training-data)

[View PolyAI](https://theagentsindex.com/polyai)