Dial gives an AI agent its own phone number. Through a REST API, CLI, remote MCP server, or Node and Python SDKs, an agent can place and receive voice calls, send and receive SMS in more than 200 countries, message through WhatsApp, and use iMessage with automatic RCS/SMS fallback. The product comes from Genway: Genway Inc, a Delaware corporation, and Genway Israel Ltd, its Israeli subsidiary. Dial’s current self-serve Terms took effect June 21, 2026.
The problem it targets is specific. Agents can execute across much of a workflow, then stall at the moment something has to happen over a phone line, like a verification code that arrives by SMS or a callback someone has to place. The telecom network still assumes a person is holding a handset with a carrier relationship behind it. Dial removes much of that friction, letting a user provision a number in seconds with $5 in signup credit and no credit card.
Dial is infrastructure rather than software your team logs into. The question worth asking is whether your agent needs a phone identity at all.
What It Does Well
Multiple channels behind one number
Dial presents voice, SMS, WhatsApp, and iMessage as channels attached to an agent’s phone identity, all reachable through a common API surface. Its published pricing table lists a unified signed webhook for calls and SMS. Its product materials describe one API, CLI, MCP server, or SDK for the platform as a whole. Calls include live transcription, and Dial documents an optional cold transfer to a human.
Assembling a comparable stack independently can require a CPaaS account, voice-agent infrastructure, an iMessage provider, and integration code to normalize events and payloads. Dial packages those communication primitives for agent-oriented workflows. That packaging, more than the ability to place a call, is what you are buying.
Setup an agent can complete
Dial publishes an agent-readable version of its site and a skills.md setup guide. During signup, its CLI can install a Dial skill into supported agent runtimes, including Claude Code, Cursor, Codex, OpenCode, Pi, OpenClaw, NanoClaw, and Hermes. A new account still requires a verified email address and a real phone number that can receive SMS. Dial instructs an agent to ask the user for that number rather than inventing or reusing one.
Agent-native onboarding at this level is still rare. A capable agent or developer can handle the CLI installation, verification, and number provisioning with little manual setup. A nontechnical business user should not mistake it for a conventional no-code application.
Your model or theirs
Managed calls use Dial’s hosted model at $0.22 per minute. In self-hosted mode, Dial streams text over a WebSocket so the customer’s own LLM can hold the conversation at $0.13 per minute. That matters if your team has already built prompts and guardrails around a preferred model.
How Each Role Uses It
Operations, most of all
RevOps and internal automation teams are the clearest fit. They often own agents that run unattended and can stall at a phone-dependent step.
An agent signs the company up for a vendor trial, receives the SMS verification code, returns the code to the workflow, and finishes the setup without anyone watching a handset. Or an inbound number answers after hours from a written instruction and routes a transcript or structured summary into your ticketing workflow.
The caveat is that the business, not Dial, stays responsible for consents, recording and messaging rules, and any required carrier registrations.
Sales, with conditions
A Dial number can place an AI-handled follow-up call to a customer who asked for one, text an opted-in prospect a rescheduling link, or confirm a demo on the morning of the meeting and return a transcript.
It is not a responsible foundation for high-volume cold outreach. Dial’s acceptable-use terms prohibit unlawful, unsolicited, or non-consented calls and messages. Its Terms also place TCPA compliance, opt-out handling, do-not-call obligations, consent revocation, and suppression-list maintenance on the customer.
A narrow lane in research
For B2B research, the defensible use is recruitment logistics: SMS reminders for scheduled respondents, no-show follow-ups, and rescheduling.
Delegating the interview itself to an AI voice agent is a separate research-design decision. It raises recording-consent, disclosure, respondent-experience, and data-quality questions. We would not make that choice lightly, particularly for qualitative work where rapport and probing quality matter.
Weaker fits
Marketing, Finance, and HR are weaker fits. Dial provides no campaign management, consent capture, or suppression-list tooling, so it is not an SMS marketing platform, and its capabilities are not designed around finance workflows or finance-system integrations. An AI voice agent contacting candidates raises candidate-experience, fairness, disclosure, and compliance questions long before anyone gets to the integration work.
Where It Could Be Better
Messaging compliance still takes time
Provisioning a number can happen in seconds. Getting approved for high-volume U.S. SMS requires 10DLC/A2P registration, an additional one-time fee, and an estimated three-to-five-business-day process according to Dial’s pricing page. Dial’s Terms make clear that consent, opt-out, do-not-call, and carrier-registration obligations remain the customer’s responsibility from the first message onward.
OTP use requires legal and policy review
The ability to receive a phone verification code programmatically is a headline use case. It can also let an agent clear a verification step that a third-party service may have designed to confirm the presence of a person or to deter automated account creation.
Many services restrict automated registration, automated verification, or use of virtual and programmable numbers in their terms. Review the target service’s terms, applicable anti-fraud rules, and internal policy with counsel before putting this workflow into production.
Self-serve commercial terms are restrictive
The self-serve Terms provide no SLA, uptime commitment, or guaranteed availability. Dial’s total liability is capped at the greater of fees paid in the prior 12 months or $100. Disputes are governed by Israeli law and fall under the exclusive jurisdiction of Tel Aviv-Jaffa courts.
Prepaid balances are non-refundable and non-withdrawable. Unused prepaid balances expire after 12 consecutive months of account inactivity.
Pricing also needs a buyer’s attention. The public pricing comparison table shows pay-as-you-go and Enterprise plans, while the Terms still describe a flat-fee subscription of $20 per number per month or $200 per number per year. Dial’s homepage calls this subscription optional and directs customers to the billing dashboard for current rates. Treat the Terms and dashboard as the final check before committing to a plan.
An iMessage-capable number is priced at $250 per month, with messages included. Dial says that iMessage numbers are pay-as-you-go only and cannot be added to a subscription.
Why Not Just Use ChatGPT, Gemini, or Claude?
A general-purpose AI assistant does not itself provide a programmable phone number, telecom routing, carrier-facing messaging registration, or a channel identity that can receive inbound calls and texts. Dial sits below the model layer. It is communication infrastructure that an AI system calls through an API, MCP server, CLI, or SDK.
The more relevant comparison is with telecom and voice-agent infrastructure. Twilio and Telnyx offer broad CPaaS capabilities. Vapi and Retell focus on voice-agent workflows, and specialist vendors handle iMessage. Dial’s practical differentiation is its attempt to bundle an AI agent’s number, voice, SMS, WhatsApp, iMessage, and agent-facing integration surfaces into one product. Judge it on the integration work it removes from your workflow rather than on the per-message or per-minute rate.
Security and Compliance
Confirmed
Dial’s Privacy Policy says it does not use call recordings, message content, or transcripts to train, fine-tune, or develop machine-learning or AI models. It states that customer communications content is processed on the customer’s behalf under its DPA.
The published privacy and DPA materials also describe:
- Encryption in transit and at rest
- Access controls, logging, monitoring, incident-response processes, and backup/restoration capabilities
- A public DPA accepted with the self-serve Terms, without a separate signature
- U.S. hosting and processing
- Standard Contractual Clauses and the UK International Data Transfer Addendum for applicable cross-border transfers
- Stripe as the independent controller for payment-card information
Questions to ask before approval
Dial’s public materials raise several due-diligence questions:
- SOC 2: Request the current SOC 2 report, report type, audit period, scope, exceptions, and applicable legal entity. Do not treat marketing language or a badge alone as equivalent to reviewing the report.
- Retention: Dial says it has no fixed deletion timelines for account records and communications data, although customers may submit deletion requests subject to legal and legitimate-business retention needs.
- Sub-processors: Dial says its specific sub-processor list is available on request rather than published publicly. The published categories include cloud infrastructure, telecommunications connectivity, speech-to-text and voice processing, logging and monitoring, and product analytics.
- Transcription: Dial identifies transcription as third-party processing but does not publicly name the transcription provider in the policy reviewed.
- EU and UK representation: Its Privacy Policy says its GDPR Article 27 representative is “to be appointed.”
- Identity and access management: The public materials reviewed do not describe SSO, SCIM, role-based access control, or audit logs for self-serve customers. Confirm availability, plan restrictions, and roadmap commitments directly with the vendor.
- Emergency calling: Dial does not support 911, E911, or other emergency services. Its Terms require customers to disclose that limitation to end users and other people using numbers through the customer’s account.
One item for IT reviewers
Dial serves an agent-readable homepage that includes setup instructions for AI agents. Its onboarding materials direct an agent to ask the human user for a real phone number needed for verification. Read that as intentional product design rather than a flaw. It is still a useful reminder that browsing agents should treat webpage text as untrusted data and should not execute instructions from a page without appropriate policy controls and user authorization.
Data and AI Connectivity
What it reaches
Dial publicly supports:
- REST API
- Remote MCP server
- CLI
- Node and Python SDKs
- Tools or integrations for LangChain, Vercel AI SDK, CrewAI, AutoGen, and Microsoft Agent Framework
- Signed webhooks for communication events
What you build yourself
The product materials reviewed do not identify native integrations for CRM, calendar, help-desk, or drive platforms. If your team wants to create a Salesforce activity from a call transcript, open a Jira issue, update a calendar booking, or save a transcript to a document repository, expect to build and maintain that workflow yourself.
That is a reasonable trade-off for a developer-oriented communications layer, and a poor one for a team hoping to buy a finished, no-code business process.
Ratings
| Dimension | Rating | Rationale |
|---|---|---|
| Usability | 3.0 / 5 | Setup is unusually agent-friendly for a developer platform, but the CLI, API, MCP server, and SDKs are the product. There is no evidence of a full nontechnical business-user interface. |
| Power | 4.0 / 5 | Voice, SMS, WhatsApp, iMessage, live transcription, cold transfer, self-hosted voice mode, and agent-oriented connectivity form a strong capability set for its stated job. |
| Flexibility | 3.0 / 5 | Broad channel and framework coverage, but a narrow communications role and no identified native business-application integrations. |
| Cost | 3.5 / 5 | A $3 monthly standard number and $0.02 U.S. SMS create a low entry cost. The $250 monthly iMessage number, non-refundable prepaid balances, and differing plan presentation across the public pricing page and Terms complicate the picture. |
Best-Fit Roles
Dial is strongest for Operations and technical Sales teams running consented, well-governed customer communications, and useful to research teams for recruitment logistics. Marketing, Finance, and HR get much less from it, since they need campaign tooling, native business-system integrations, or workflow-specific controls.
Conclusion
Dial does one job cleanly: it gives an AI agent a programmable identity on the phone network. It does that through a broad set of communications channels and agent-facing integration options, with quick self-serve provisioning and a low-cost entry point for a standard number.
If an agent is blocked by a verification code, callback, inbound phone request, or opted-in follow-up, Dial may remove real integration work. If the workflow does not need a phone identity, there is little reason to adopt it.
Human judgment stays load-bearing in two places. Someone has to decide what the agent is permitted to say to a real person, and someone has to own the consent, recording, disclosure, carrier-registration, and opt-out obligations that come with a phone number. Dial makes the call technically possible. Whether it should be made is still a person’s decision.
Last updated: September 4, 2026.