Voice agents and telephony APIs and MCP servers for AI agents
Place and receive phone calls handled by AI voice agents, real-time audio/video rooms.
8 voice agents and telephony APIs an AI agent can call: auth method and header, official MCP server (7), free tier (3), docs links. Machine-readable JSON included. Agents: fetch catalog/voice-agents-telephony.json instead of parsing this page. Last checked 2026-09-24.
Ratings and usage (★) fetched: .
| Service | What an agent can do | Auth | MCP server | Free tier | Details (with sources) | Notes | Proof (ratings, usage) |
|---|---|---|---|---|---|---|---|
| Bland AI | Programmatically create AI phone agents and place/manage automated outbound and inbound calls. | API keyAuthorization: Bearer <api_key> header (bare key without prefix still accepted) | official | free credits + inbound number on signup (Start plan), no credit card required | Base URL: https://api.bland.ai/v1 Free plan: 2 credits + an inbound number; 100 calls/day, 10 concurrent calls, no card ↗ Rate limits: Start plan: 100 calls/day, 100 calls/hour, 10 concurrent calls; MCP server 120 req/min per org ↗ Async jobs: POST /v1/calls returns call_id; poll call status/logs or pass webhook to receive the call object when it ends ↗ Example call: place an outbound AI phone call: POST https://api.bland.ai/v1/calls ↗ Unit price: Connected calls: $0.14/min on Start (no platform fee), $0.12/min on Build ($299/mo platform fee) ↗ MCP tools: 13: create_call, wait_for_call, stop_call, get_call_log, list_voices, validate_pathway … ↗ | MCP server rate-limited to 120 req/min/org; distinguishes read-only vs destructive operations. | — |
| Deepgram Voice Agent API | Run a full-duplex speech-to-speech voice agent pipeline (STT+LLM+TTS) over a WebSocket API. | API keyAuthorization: Token <api_key> | official | $200 free credit on signup | Free plan: $200 credit, one-time, no card ↗ Rate limits: Pay As You Go: up to 45 concurrent Voice Agent WSS connections (60 on Growth); STT 150 WSS / 50 REST; TTS 45 ↗ Data policy: Model Improvement Program may keep data for training; opt out per request with mip_opt_out=true (data kept only for processing) ↗ Unit price: Voice Agent Standard: $0.075 per minute of WebSocket connection (Pay As You Go); BYO TTS $0.065/min, BYO LLM $0.059/min, BYO LLM+TTS $0.050/min ↗ | Primarily an STT/TTS/voice-agent infra layer, often paired with a telephony provider like Twilio to reach phone calls. MCP server built into the dg CLI ('dg mcp', stdio or SSE); docs MCP at https://developers.deepgram.com/_mcp/server. | — |
| ElevenLabs Agents Platform | Build and deploy conversational AI voice agents with telephony (SIP/Twilio) integration via API and SDKs. | API keyxi-api-key: <api_key> | official | Free plan: 10k credits/month shared across ElevenLabs products; no commercial license (from Starter, $6/month) | Base URL: https://api.elevenlabs.io/v1 Free plan: 10k credits, per month ↗ Data policy: Zero Retention Mode available per agent and for Enterprise API use ↗ Example call: create a conversational agent: POST https://api.elevenlabs.io/v1/convai/agents/create ↗ Unit price: Agents (Speech Engine): $0.08 per minute beyond plan-included minutes, $0.16 per burst minute ↗ | Official hosted MCP server uses OAuth (sign in with ElevenLabs account) rather than a raw API key; core REST API uses api-key auth. | — |
| LiveKit Cloud | Build and deploy real-time voice AI agents with SIP telephony support using LiveKit's server SDKs and APIs. | API keyAPI key + secret sign a JWT access token sent as Authorization: Bearer <token> | — | free Build plan: 1,000 agent session minutes/mo, 5,000 WebRTC participant minutes/mo, 50GB data transfer | Free plan: 1,000 agent session min, 5,000 WebRTC participant min, 50GB transfer, $2.50 inference credits, per month, no card ↗ Rate limits: Build plan: up to 100 concurrent users and agents ↗ Example call: dispatch a voice agent into a room: POST https://{projectDomain}/twirp/livekit.AgentDispatchService/CreateDispatch ↗ Unit price: Agent sessions: $0.01 per minute overage on paid plans (Build includes 1,000 min/mo free); WebRTC overage $0.0004-$0.0005 per participant minute ↗ | Self-hosted open-source option also exists; Cloud is the managed offering with usage-based billing beyond the free tier. | — |
| Retell AI | Build, deploy, and manage AI voice agents that handle phone and web calls via REST API and SDKs. | API keyAuthorization: Bearer <api_key> | official | $10 free credit on signup, then pay-as-you-go ($0.07-$0.31/min for voice) | Base URL: https://api.retellai.com Free plan: $10 in free credits, one-time ↗ Rate limits: Pay-As-You-Go workspaces: 20 concurrent calls by default (extra $8/call/month); calls max 1 hour by default ↗ Example call: place an outbound AI phone call: POST https://api.retellai.com/v2/create-phone-call ↗ Unit price: Per call minute: voice infra $0.055 + LLM (e.g. $0.064 GPT 5.6 Terra) + telephony ($0.015 US via Twilio); advertised $0.07-$0.31/min ↗ | Official MCP at https://mcp.retellai.com (Streamable HTTP) authenticates with Authorization: Bearer <RETELL_API_KEY>. | SourceForge 1.0★/1 review |
| Synthflow API | Create and manage AI phone voice agents and trigger/manage outbound calls programmatically via REST API. | API keyAuthorization: Bearer <api_key> | official | Paid only: pricing page lists Enterprise contracts from $30,000/year; no self-serve free trial shown | Base URL: https://api.synthflow.ai/v2 Rate limits: Per-workspace concurrency limit and CAPS (calls initiated per second) limit; excess returns 429 or is queued ↗ MCP tools: 17: list_agents, get_agent, create_agent, update_agent, publish_agent_draft, delete_agent … ↗ | Go-to-market is enterprise sales. Official remote MCP (~60 tools) also has regional URLs: https://mcp.us.synthflow.ai/mcp and https://mcp.eu.synthflow.ai/mcp. | SourceForge 1.0★/1 review |
| Twilio Programmable Voice | Programmatically make, receive, and control phone calls (IVR, recording, conferencing) via REST API. | API keyBasic Auth (Account SID + Auth Token) or API Key/Secret | official | Trial: one-time free units (e.g. 75 voice minutes, 100 SMS, 100 WhatsApp, 3,000 emails), no card, expires after 30 days; then pay-as-you-go | Base URL: https://api.twilio.com/2010-04-01 Free plan: 75 voice minutes, 100 SMS, 100 WhatsApp messages, 3,000 emails; verified numbers only, one-time, no card ↗ Rate limits: Account API concurrency limit; exceeding returns 429 Too Many Requests ↗ Example call: make an outbound call: POST https://api.twilio.com/2010-04-01/Accounts/{AccountSid}/Calls.json ↗ Unit price: US: outbound calls $0.0140/min; inbound $0.0085/min on local numbers (+$1.15/mo per number) ↗ | Widely used as the underlying telephony carrier beneath voice-agent platforms (Vapi, Retell, ElevenLabs, LiveKit) for actual call carriage. Twilio's own MCP (npm @twilio-alpha/mcp) is alpha. | GitHub 169★ twilio/twilio-oai GitHub 111★ twilio-labs/mcp |
| Vapi | Create, configure, and run AI voice assistants that make and receive phone calls via REST API. | API keyAuthorization: Bearer <api_key> | official | $5 free credit on signup, then pay-as-you-go (~$0.05/min platform fee + provider costs) | Base URL: https://api.vapi.ai Free plan: $5 free credits, one-time ↗ Rate limits: 4 concurrent calls with no Success Package; 10 on Core ($29/mo), 30 on Pro ↗ Example call: start an outbound phone call with an assistant: POST https://api.vapi.ai/call ↗ Unit price: $0.05 per call minute Vapi hosting fee; transcriber, LLM, voice and telephony billed separately at cost ↗ MCP tools: 10: list_assistants, create_assistant, get_assistant, list_calls, create_call, get_call … ↗ | Also has an official MCP server (streamable-HTTP and SSE endpoints) exposing assistants/calls/phone numbers as tools. | — |
Other categories
- Web search
- Web scraping and browser automation
- Knowledge and research data
- Image generation and editing
- Video generation and editing
- Speech and audio
- Music generation
- 3D generation and assets
- Architecture, CAD and BIM
- Diagrams and software architecture
- Design and UI
- Documents, OCR and presentations
- Translation and language
- Code execution sandboxes
- Developer platforms and DevOps
- Cloud and hosting
- Databases
- Vector search, memory and RAG
- Email, team chat and notifications
- SMS and messaging apps
- Productivity and workspace
- CRM, support and marketing
- Social media
- Maps, geolocation and weather
- Finance, payments and market data
- E-commerce
- Automation and integration platforms
- Model APIs and inference