Voice agent buyer guide
AI Voice Agent Pricing: Calculate the Real Cost per Call
Normalize platform, model, speech, carrier, phone-number, transfer, recording, storage, and implementation charges against one call workload.
What the terms mean
Start with monthly connected minutes, attempts, average call length, peak concurrency, countries, transfers, recordings, and actions. Ask each vendor to price that identical workload.
- Evidence
- A primary document, observed pilot result, or contract term that supports a claim.
- Production path
- The complete route from caller and carrier through models, tools, transfer, logs, and human follow-up.
Decision record
Compare each option across the same operational criteria.
Swipe or scroll the table to see every column.
| Criterion | Input or need | Evaluation scope | Evidence to collect |
|---|---|---|---|
| Buyer requirement | A feature claim | A pass/fail production behavior | Recorded scenario and artifact |
| Cost | Headline rate | Complete workload cost | Quote with every component |
| Risk | Policy statement | Tested control and owner | Configuration, log, and contract |
Reusable decision tool
Build the evidence pack
Check the work that must be complete before a decision.
Bring your own quote · no signup
Compare two quotes against one workload
Enter the billable units each vendor says the same workload produces. This tool does not turn conversations into resolutions or predict an automation rate. Your inputs stay in this page and are cleared on reload.
Monthly = other monthly costs + max(0, units − recurring included units) × unit rate.
Year one = 12 × monthly + one-time costs. Usage rounds to cents each month.
Planning estimate only, not a vendor quote or savings forecast. Assumes identical monthly usage and a flat unit rate. No automatic conversion between conversations, outcomes, credits, actions or minutes. Enter taxes and all other charges yourself. Tiered rates, minimum commitments, prepaid packs, ramp-up and one-time promotional credits require a separate model. Two totals are comparable only if scope, unit definitions and cost coverage have been checked.
Complete the workload, both vendor names, unit definitions, all numeric inputs and assumptions to export.
Calculate cost per minute, call and completed task
Effective cost per minute is the complete monthly voice bill divided by connected minutes. Cost per call divides that bill by connected calls. Cost per completed task divides it by calls that achieved the agreed outcome. State whether engineering, support and human follow-up are included.
Illustrative calculation, not a vendor quote: 2,000 connected calls averaging three minutes produce 6,000 minutes. A hypothetical all-in bill of $900 is $0.15 per minute and $0.45 per call. If 1,200 calls complete the intended task, the cost is $0.75 per completed task. Add any excluded implementation or human follow-up costs separately.
Request base, expected and peak quotes with the same assumptions. Separate subscription allowances, usage, carrier charges and overages so a low headline minute rate does not hide the cost of the full workflow.
Use one workload, not one advertised minute
Start with monthly connected minutes, attempts, average call length, peak concurrency, countries, transfers, recordings, and actions. Ask each vendor to price that identical workload.
Do not compare an infrastructure minute with a managed resolution or subscription allowance. Preserve every vendor's billing definition.
Price architecture and operations together
A configurable cascade can create separate STT, LLM, TTS, orchestration, and telephony charges. Native audio models use different token or minute units. Both still need carrier, tools, logging, evaluation, and human escalation.
Model base, expected, and peak months, including retries, voicemail, silence, failed calls, and transferred call legs.
Take this to every provider
Your evaluation checklist
- Export monthly attempts, connected minutes, and concurrency
- Include all model and speech components
- Include carrier, number, SIP, and transfer charges
- Separate implementation from recurring operations
- Model base, expected, and peak usage
Use the decision tool above and the full pilot worksheet to record results against each shortlisted provider: Shortlist workspace.
Primary sources
Source summaries reviewed 2026-09-04.
Vapi lists a hosting fee while model, voice, transcriber, and telephony costs pass through.
Retell itemizes infrastructure and optional call components.
Twilio prices ConversationRelay separately from Programmable Voice.
Providers to investigate
See all 27 providers →These examples span the operating models covered in this guide.
Swipe or scroll the table to see every column.
| Providers | What to investigate | Pricing model |
|---|---|---|
| Vapi | Verify Configurable pipeline, including STT-LLM-TTS and OpenAI Realtime on your real phone route; then test Not publicly documented. | Usage-based hosting plus model and telephony pass-through; Scale adds an annual contract, committed volume, and volume pricing. Profile sources · checked 2026-09-04 |
| Retell AI | Verify Configurable STT-LLM-TTS pipeline with custom LLM option on your real phone route; then test Not publicly documented. | Pay-as-you-go per-minute component pricing with optional add-ons; Enterprise uses custom volume pricing and terms. Profile sources · checked 2026-09-04 |
| ElevenLabs Agents | Verify Cascaded STT-LLM-TTS using ElevenLabs voices on your real phone route; then test Configurable turn-taking, interruptions and timeouts and Human transfer with channel-specific conference, blind and SIP REFER modes. | Monthly plans include call minutes and concurrency. LLM usage is billed separately; external telephony providers bill their own carrier charges. Enterprise terms are custom. Profile sources · checked 2026-10-09 |
| ConversationRelay | Verify Cascaded speech interface with developer-supplied LLM on your real phone route; then test Not publicly documented. | Usage-based Programmable Voice minutes and phone numbers, plus ConversationRelay per-minute charges; model, provider, and other call features are separate. Profile sources · checked 2026-09-04 |
| Goodcall AI Phone Agent | Verify Not publicly documented on your real phone route; then test Not publicly documented. | Per-agent monthly subscription with unique-customer allowances and overage charges; Enterprise is custom. Profile sources · checked 2026-09-04 |
Before you decide
Questions buyers ask
What does an AI voice agent cost per minute?
There is no universal rate. Vendors may bill hosting, STT, LLM, TTS, native audio tokens, telephony, numbers, transfers, recording, storage, quality analysis, and support separately. Calculate the full stack for your workload.
Are unlimited-minute plans really unlimited?
Treat unlimited as a contract term, not a universal promise. Check fair-use rules, concurrency, included numbers, premium models, transfers, and prohibited workloads.
Keep building your shortlist
Browse voice-agent providers →Compare candidates →Build an evidence-backed shortlist →All buying guides →Evaluate language support in real calls
Test recognition, speech quality, interruptions, language switching, tool use, and human transfer with your own calls in the required language. Availability can vary by model, plan, and phone route.
DocsBot documents multilingual voice workflows and Japanese support. Verify the exact voice, phone path, and transfer behavior in your pilot. Review DocsBot language information ↗
Other providers’ language claims reflect cited vendor material; confirm Japanese and other required languages directly with each provider.
Listing does not imply endorsement or a performance guarantee. This directory is published by DocsBot. Editorial policy