ThunderPhone
Build AI voice agents that handle real phone calls, from 2 cents a minute.
Overview
ThunderPhone is a platform for building AI voice agents that handle real, live phone calls rather than just text or in-app chat. You write a prompt, choose a voice and an "intelligence" tier (Spark, Bolt or Storm), and the agent can run inbound support lines or outbound calling campaigns in more than 40 languages, with interruption handling and mid-call transfers built in. Pricing undercuts the competitors it names: ThunderPhone quotes roughly 2 cents a minute at its cheapest against a stated 7 cents or more elsewhere, and well below human call center rates.
This is a genuinely early-stage product. ThunderPhone originally launched in March 2026; its self-serve "v2" platform only went publicly live on August 27, 2026, about two weeks before this listing, after what the founder's own blog post called "a multi-month odyssey" past the original limited-access period. Some third-party trackers still describe it as being in alpha, though the vendor's current site language calls this a "public launch" rather than an alpha. Either way, expect documentation, edge cases and support to still be maturing.
Key Features
Inbound and outbound calling, including campaigns with CSV upload (up to 5,000 contacts)
Three "intelligence" tiers - Spark, Bolt, Storm - trading cost against capability
Multi-path audio transcription, described by the vendor as fusing multiple transcripts with direct audio-to-LLM input
Vendor-reported 99.4% accuracy on the Big Bench Audio benchmark (July 2026) - not independently verified
Handles interruptions, backchannel sounds ("mm-hm") and mid-call language switching
Knowledge base retrieval, warm/cold call transfers, voicemail detection
Live mid-call supervision and guidance
No-code builder plus API access; can import existing agents from Vapi, Retell, ElevenLabs and Bland
40+ supported languages with accent recognition
Pricing
Starting price
Usage-based, from $0.02/min (Spark tier) - no monthly or annual plan published
Usage-based, no subscription. Three "intelligence" tiers, billed per minute: Spark $0.02/min, Bolt $0.05/min, Storm $0.09/min. Add-ons: premium voices +$0.03/min; prompts over 5,000 tokens add a per-10K-token-block surcharge (+$0.01/min on Spark, +$0.02/min on Bolt and Storm). Phone numbers cost $1/month plus $0.01/min for inbound calls. No flat enterprise or volume pricing is published; compliance packages (HIPAA BAA, GDPR DPA) are enterprise-only with pricing not disclosed on the site. Source: thunderphone.com/pricing and thunderphone.com/docs, checked 2026-09-13.
Pros
Meaningfully cheaper per minute than the competitors it names (2c/min vs a stated 7c/min+), and well below human call center rates
Real technical substance behind the pitch: multi-path transcription, interruption handling, mid-call language switching, and a published (if vendor-reported) benchmark score
Can import agents built on other platforms (Vapi, Retell, ElevenLabs, Bland), which lowers switching cost for teams already building voice agents elsewhere
Covers both non-technical and developer use cases with a no-code builder alongside a full API
Cons
Early-stage product: the self-serve v2 platform only went publicly live on Aug 27, 2026, weeks before this listing, and the vendor's own blog describes the runway to get there as an unexpectedly long "multi-month odyssey" - expect rough edges
No affiliate or partner program found on the site
Pricing beyond the base per-minute rate is not fully published for every scenario. Confirmed add-ons are +$0.03/min for premium voices and a +$0.01-0.02/min surcharge for prompts over 5,000 tokens, plus $1/month + $0.01/min per phone number - but no enterprise or volume pricing is public
Compliance features (HIPAA BAA, GDPR DPA) are enterprise-only with pricing not disclosed; TCPA-compliant cold outbound requires a separate "ThunderHuman" human-agent service, not the core AI product
Very limited independent, long-term user reviews given how recently v2 launched
What Makes It Unique
ThunderPhone's angle is a meaningfully lower per-minute price than the voice-agent platforms it names as competitors, paired with the ability to import agents already built on those competitors (Vapi, Retell, ElevenLabs, Bland) instead of requiring a rebuild from scratch.
Kay Score
5
/ 10
Tool Information
Pricing
Usage-based, from $0.02/min (Spark tier) - no monthly or annual plan published
Category
Automation & AI Agents
Platform
Web / iOS / Android
Last Updated
