Niches · Conversational voice agents
Dial a list and qualify, remind, collect or book
No company's docs show a full fit yet. The listings below do the job with a limit, or not at all.
What counts as done
Each contact is dialled once (idempotent), voicemail is detected and either skipped or a message left, every connected call ends with a structured outcome and transcript, and cost per completed call is readable from the log.
Measured, for our own tests: a voicemail message is left in under 20 seconds. No docs we read state these figures, so they don't count against a listing here.
How agents ask for it: call these 500 leads and book a demo
, remind tomorrow's patients
, chase overdue invoices by phone
Does the job, with a limit 13
It does the job, but something in the docs or a user's account will get in an agent's way. The limit is named in each row.
| Listing | Why |
|---|---|
| Bland AI API + MCP B Intelliga Corporation | Voicemail detection (default hangup, leave_message, retry), batch calls that drop duplicates and do-not-call numbers at dial time. No idempotency key on POST /v1/calls, no calling-window check found, and failed attempts are billed. source |
| Bolna API + MCP D Voxlabs Private Limited | Batches, voicemail detection, an outbound calling window that reschedules into the recipient's timezone, and a cost_breakdown per execution. Voicemail detection appears off in the example config and the window is off by default; no idempotency key or DND registry check. source |
| ElevenLabs Agents API + MCP BB Eleven Labs Inc. | Campaigns add calling windows by timezone, retries and callbacks; conversations return charging and cost fields. Voicemail detection is a system tool that must be added to API-created agents, and the outbound call endpoints show no idempotency key. source |
| Harmony.ai E monday.com Ltd. | Imported contacts are called by workflow within working hours in each contact's timezone, with attempt numbers and outcomes including voicemail, do_not_call and dnc_blocked. No idempotency key (a retried enrolling import can duplicate) and no cost field per call. source |
| LiveKit Agents BB LiveKit Incorporated | Answering machine detection classifies voicemail, but it must be set up per call; outbound SIP has no idempotency key, no calling-window or DND check, and session.usage gives usage counts rather than cost per call. source |
| Pipecat B Daily | VoicemailDetector and Daily dial-out are documented, but there is no idempotency key on dial-out, no calling-window or DND check, and usage metrics report tokens and characters per service rather than a cost per call. source |
| PlaceCall C VOYGR Tech, Inc. | Idempotency-Key stops double dials, voicemail returns as a free failure reason and each call gives an outcome, transcript and flat 10-credit cost; but it targets US businesses only, excludes five states, and leaves no voicemail message. source |
| Retell AI API + MCP B Retell AI Inc. | idempotency_key on create-phone-call, Dialer calling windows and Retell Contacts DNC, voicemail hang-up or message, and call_cost with product_costs per call; but API-created agents must set voicemail_option explicitly, TCPA hours are not enforced, and spam-flagged numbers return dial_failed. source |
| Synthflow API + MCP D AgentFlow AI GmbH | Campaigns take up to 10,000 recipients, call_time_window per weekday, and a recipient id that skips duplicates on retry. Leaving a voicemail needs Human Speaks First, there is no DND registry check, and per-call cost was not found in the docs read. source |
| Vaani Developer Platform E Vaani AI Research Private Limited (Bengaluru, India, CIN U62099KA2025PTC201212) | Campaigns enforce the TRAI 08:00 to 19:59 IST window, offer perform_dnd_check (default false), voicemail detection with hang_up or leave_message, retries and a call_postprocessing webhook with transcript and dispositions. No idempotency key on call creation, and cost is one credit figure. source |
| Vapi API + MCP B Vapi Inc. | Campaigns with per-contact variables, a scheduled window, a pre-dial eligibility webhook and voicemail detection with a message, and a per-call costBreakdown. Voicemail detection is off by default, POST /call has no idempotency key, and no DND registry check. source |
| Vocily AI C VOCILY AI PRIVATE LIMITED | POST /v1/calls requires an Idempotency-Key, voicemail is detected and recorded with voicemail_action hang_up or leave_message, campaigns have retry rules, a call window and duplicate removal, and each call carries cost.total. voicemail_action defaults to none and no DND registry check exists. source |
| Vogent API E Monoid, Inc. | POST /dials takes an idempotencyKey (reuse returns 409), voicemailConfig has HANGUP or LEAVE_VOICEMAIL, batch dial jobs run in per-day time slots with a timezone, and transcripts and extractors arrive by webhook. No DND check and no per-call cost field in the API. source |
Couldn't check 1
The docs were closed to us, or silent. These carry no mark either way.
| Listing | Why |
|---|---|
| Hume EVI (Empathic Voice Interface) C Hume AI, Inc. | EVI docs withdrawn: every EVI page, including the overview and tool-use guide, redirects to the Audio Decisions overview, and the EVI AsyncAPI file returns 404. The last assessment recorded EVI access ending 13 November 2026. source |
Doesn't do this job 13
Listings in the category that do something else, with the reason.
| Listing | Why |
|---|---|
| Amazon Nova 2 Sonic B Amazon Web Services, Inc. | A model API with no dialling, numbers, voicemail detection or campaign functions; all of that sits in the bridged carrier. source |
| AssemblyAI Voice Agent API C AssemblyAI, Inc. | Telephony is inbound only over SIP; no outbound dialling, voicemail detection or campaign API. source |
| CueLight D sshah03 (individual GitHub account, no name or organisation stated) | CueLight is an end-of-turn detector only: no call handling, telephony, tools or speech output. source |
| Deepgram Voice Agent API BB Deepgram, Inc. | Outbound calling is a Twilio or Plivo tutorial the developer runs; no dialler, voicemail detection, numbers or campaign controls in the API. source |
| fone.me F Fone | Calls are one-off errands for the owner from Premium minutes; no list dialling, call API or campaign controls are offered. source |
| Fresha AI Concierge F Fresha.com SV Ltd | Answers inbound calls and messages only; no outbound dialling is described. source |
| Gemini Live API C Google LLC | A model API with no dialling, numbers, voicemail detection or campaign functions. source |
| GPT-Live BB OpenAI OpCo, LLC | Outbound SIP places one call per session create, with no idempotency key (a retry after timeout can dial twice), no voicemail detection and no list, window or DND functions. source |
| Inworld Realtime API B Theai, Inc. dba Inworld AI | No outbound dialling, numbers, voicemail detection or campaign functions; telephony is an inbound Twilio bridge example. source |
| OpenAI Realtime API BB OpenAI OpCo, LLC | The Realtime SIP guide covers accepting inbound calls only; outbound dialling, voicemail handling and campaign functions are documented for GPT-Live, a separate product. source |
| Pipecat PhoneLLM B Daily | No dialling, voicemail or campaign functions; a text model only. source |
| Rime Prism B Rime Labs, Inc. | No telephony, dialling or voicemail functions; Prism is a WebSocket speech-to-speech model. source |
| Ultravox Realtime API C Theai, Inc. dba Inworld AI | The outbound scheduler handles windows, concurrency and retries over the developer's carrier, but the API schema has no voicemail detection and no idempotency key, so the success criterion cannot be met from the platform. source |
What stops agents finishing
- voicemail detection off by default for API-created agents
- outbound number flagged as spam by carriers after many short calls
- failed attempts billed
- TCPA, TRAI or DND rules with no platform check
Around it
Before: crm.export-list, scheduling.calling-window. After: crm.update-record, booking.reservations, analytics.call-outcomes.
Where this comes from
Weight 14 of 100 in Conversational voice agents: 5 of 31 accounts (acc-12, acc-13, acc-18, acc-29, acc-30); the ventureharbour test built outbound appointment setting on 5 platforms
The same as data: /api/v1/use-cases/voice-agents.outbound-campaign.json, or the MCP tool get_use_case with id voice-agents.outbound-campaign. Other jobs in this category: /api/v1/niches?category=voice-agents.
For companies
Do agents find, use and choose your tools?
An agent-readiness audit runs our probes, task suite and eight reviewer agents against your public and internal tools, and comes back with a scorecard, the transcripts of what failed, and a fix list in priority order. From $2,500, re-run included. We never take payment to move a rank. We do help companies earn one.