TL;DR
- Twilio is the strongest default for an agency building client-specific calling applications. Subaccounts separate resources and credentials, while the agency pays one parent bill.
- Plivo is a good lower-cost alternative for straightforward US calling. Its subaccounts have separate credentials, logs and webhooks; its documented number-porting service covers the US and Canada.
- Telnyx suits teams combining number inventory, SIP and programmable voice. Budget both its Voice API charge and carrier minutes. Managed Accounts require a committed plan, so low pay-as-you-go call costs do not establish the cost of client account management.
- Bandwidth fits a communications platform managing larger inventories. Sites and locations organise numbers and routing, but are not a promise of independently billed client accounts.
- Vonage Communications APIs makes sense for an existing Vonage API stack with the required subaccount access. Dial is a specialised US agent-number option, best provisioned in the client's own account rather than assumed to be an agency resale platform.
- Preserve the client's callback number at campaign end. Stop outbound activity, transfer or port the retained number, then close the old account. Releasing a number is not a migration method.
What the agency is actually buying
Phone-number management becomes a buying decision when an agency has several clients, several campaigns per client and an application that provisions numbers automatically. The difficult part is keeping each client's numbers, routing, documents and bills together as campaigns change.
A useful provider gives you searchable inventory, number-specific capabilities, programmable assignment, client separation and a workable exit. A cheap number without those capabilities can cost more to operate than a slightly more expensive number in an established account model.
Separate three decisions. The number provides an address people can call back. The calling service connects and bills call legs. The application supplies a dialler, receptionist, CRM update or AI conversation. Buying a number does not include every application feature, and an AI-agent minute is not comparable with a bare carrier minute.
For a small agency, one client-specific subaccount with several labelled campaign numbers is usually easier to operate than a new subaccount for every campaign. Put a separate production number on an enduring customer-facing activity. Avoid making a temporary campaign number the client's main published line unless the client will retain it.
This guide concerns the number lifecycle and its economics. Our conversation intelligence comparison addresses the separate decision about reviewing and analysing calls.
The provider comparison
Public pricing and documentation assessed on 1 October 2026. USD rates below are US local-number and calling bases, before taxes, selected extras and destination-specific differences.
| Provider | Client organisation | Useful price basis | Main limitation for this purchase |
|---|---|---|---|
| Twilio | Subaccounts with distinct credentials and resources; parent billing | $1.15/local number/month; $0.014/outbound minute; $0.0085/inbound minute | Parent suspension affects subaccounts; closing a subaccount releases its remaining numbers |
| Plivo | Subaccounts with separate credentials, logs and webhooks; shared parent balance | $0.50/local number/month; $0.0115/outbound minute; $0.0055/inbound minute | Published porting coverage is US/Canada, narrower than its new-number inventory |
| Telnyx | Managed Accounts with their own account settings; rollup billing option | Local numbers from $1/month; Voice API $0.002/minute plus SIP rates | Managed Accounts documentation requires a commit plan; a connection/profile alone is not an independent client account |
| Bandwidth | Account, subaccount/site, location and application hierarchy | US Voice API $0.0100/outbound and $0.0055/inbound minute; add contractual number rental | Inventory grouping and application routing need explicit client-access and billing design |
| Vonage Communications APIs | Restricted-availability Subaccounts API; shared or individual balances | Country/destination voice rate plus number rental and selected features | Do not assume self-serve subaccount access or substitute Vonage Business Communications seat prices |
| Dial | Number IDs, nicknames, calling settings and agent instructions | $3/standard number/month; $0.13/self-hosted or $0.22/managed call minute | New numbers currently US-only; self-serve terms restrict third-party resale and do not guarantee porting |
The closest small-agency comparison is Twilio versus Plivo. Telnyx becomes particularly attractive when existing spend supports its account-management commitment. Bandwidth deserves a platform-level comparison; Dial deserves an agent-application comparison.
Six options and their strongest use cases
1. Twilio: the default for a custom agency application
Twilio combines number provisioning with a well-documented parent/subaccount model. Give a client's application its own subaccount credentials instead of the parent's credentials. This keeps the client's resources separate while making consolidated billing manageable. The parent still carries the commercial responsibility, so an agency payment failure can affect several clients.
Its subaccount documentation supports moving numbers between a parent and its subaccounts. That is useful when a campaign closes but the callback number must stay active. A move can require routing reconfiguration and destination-account compliance records; associated messaging registrations do not simply become portable configuration.
The US Programmable Voice rates are $1.15/month for a local number, $0.014/minute outbound and $0.0085/minute inbound. Toll-free rental and inbound calls use different rates. Recording, browser connections and other selected features also have their own charges.
Choose Twilio when developers already know its voice stack, clients need clear resource separation and the agency will own the application. Choose another route when the buyer wants a ready-made team phone system or independent client billing without the agency carrying a shared balance. A number API is not a finished sales dialler.
2. Plivo: straightforward calling with inexpensive US number rental
Plivo's subaccounts provide their own Auth ID/token, call and message logs, and webhooks. Charges deduct from the main account, which suits agency rebilling. Assign a number when renting it or move it later through the account number API.
The current US Voice API card lists $0.50/month local rental, $0.0115/minute outbound to the main US network group and $0.0055/minute local inbound. Alaska and Hawaii have different outbound rates. Its cheaper SIP-trunk comparison is a different product basis; do not apply those trunk rates to a Programmable Voice budget.
Plivo is a compelling choice for several modest-volume US client campaigns using a fairly simple calling application. Saving on number rental is especially useful when many numbers remain active to receive occasional callbacks.
Its published porting guide covers US and Canadian local/toll-free numbers and describes free port-in, potentially taking up to four weeks. That makes it a weaker default for migrating an existing multinational number estate, even where new-number rental is available. Preserve the distinction between buying a new number and moving an existing one.
3. Telnyx: number inventory plus carrier and voice primitives
Telnyx combines numbers, SIP trunking and programmable calling on its own platform. Its number card starts local rental at $1/month, with an additional $0.10/month for SMS/MMS capabilities. Its Voice API charges $0.002/minute in addition to the underlying SIP charge. The US SIP basis starts at $0.005/minute outbound and $0.0032/minute local inbound, producing voice bases from $0.007 and $0.0052 respectively.
The important agency feature is Managed Accounts. Current feature-specific guidance says they are unavailable on pay-as-you-go and included in Starter at a $1,000 monthly minimum with quarterly billing, Growth at $2,000, or Enterprise. That more specific requirement governs this recommendation, despite broader pricing-page language about ungated primitives and commitments from $500.
Managed Accounts support customer-specific configuration and a rollup-billing choice. Set that commercial model at creation; the documented rollup setting cannot later be changed. Telnyx also provides porting automation across more than 50 countries, useful for a broader client estate.
Choose Telnyx when the agency already has sufficient aggregate communications spend or works in client-owned accounts. Do not buy a large commitment just to obtain Managed Accounts for a small number-rental bill. API connections and messaging profiles help organise traffic but do not supply the same tenant boundary.
4. Bandwidth: a larger inventory and communications-platform purchase
Bandwidth's number-management product covers ordering and porting across an inventory advertised in roughly 70 countries. Its account structure places numbers in a subaccount/site and location/SIP peer. Numbers inherit location settings, and applications determine callback routing.
That hierarchy is useful for a platform operator: one client can have distinct locations for different routing or messaging behaviour. Applications sit at account level, so a neat inventory hierarchy does not by itself establish independently credentialled, independently billed client accounts. Design the application's client permissions and invoice attribution explicitly.
The US pricing page lists Voice API rates of $0.0100/minute outbound and $0.0055/minute inbound, with recording at $0.002/minute and other primitives separately priced. Number rental belongs in the commercial schedule; the model below leaves it visible rather than inventing an amount.
Bandwidth is attractive for a communications product or substantial estate that needs systematic inventory and porting. It is less compelling for an agency seeking the quickest self-serve setup for three small clients. Its current bulk account-move guidance covers NANP numbers within the same Global Account Number, not every international migration.
5. Vonage Communications APIs: retain it when the existing stack fits
Vonage's Numbers API provides rental and configuration through API, CLI and dashboard surfaces. Pair it with Voice API applications rather than treating it as a business-phone seat subscription.
Its Subaccounts API can create and suspend client accounts and allocate shared or individual balances. However, the documentation still marks it beta with restricted availability; partners receive access automatically. The balance choice matters because an account created with an individual balance cannot later be converted to shared balance.
Vonage supports number porting in North America and selected other markets. Its voice pricing uses country/destination rate sheets. Calculate rental plus inbound/outbound minutes and the chosen extras from that basis; there is no need to invent a global monthly package.
Keep Vonage when an existing API deployment and account arrangement already serve the clients well. For a new small-agency build, Twilio or Plivo provides a clearer starting point. That is an operational-fit judgement, not a claim that Vonage lacks global numbers or programmable voice.
6. Dial: a client-owned number for an AI agent
Dial's number documentation exposes a number ID, country, capabilities, nickname and inbound agent instruction. You can change the instruction, turn calling off or forward inbound calls to another phone. Those are useful controls for a receptionist or task-specific agent.
Provisioning is currently US-only. International SMS destinations do not mean international local-number inventory. Its pay-as-you-go pricing is $3/month for a standard number, $0.13/minute for self-hosted calls using your own model or $0.22/minute managed. These are agent-call prices, not a bare-carrier alternative to a one-cent minute.
The decisive limitation for agencies is its self-serve terms: third-party resale requires express permission, and porting is conditional rather than guaranteed. Use a client's own account for a suitable US agent project. Prefer the fuller account and porting models above for an agency's shared number estate. This preserves Dial's useful specialist role without assuming an undocumented reseller or subaccount capability.
Country inventory and existing numbers
Build the shortlist around country, number type and capability, not an advertised country count. A London local callback number, a UK mobile SMS sender and a US toll-free support number are different purchases. Outbound calling to a country does not prove local rental there; local rental does not prove SMS or porting support.
Consider a fictional agency with two US clients and one UK client. It needs six US local numbers, one new UK geographic number and a port of the UK client's long-established published number. Twilio's UK number requirements and UK porting guide provide a direct route for the UK evaluation. Telnyx has UK DID requirements and broader porting coverage. Plivo has a UK new-number catalogue, but its published porting guide does not establish that it can take the existing UK number.
My choice for this mixed estate is Twilio, unless an existing Telnyx account arrangement already makes the commitment sensible. For the US-only portion, Plivo remains a strong price-focused option. Dial drops out of the UK local-number requirement. Keeping the incumbent UK carrier is also sensible if moving the number contributes little to the campaign.
Treat the client as the documented end user. Store the provider's approved identity/address record against the actual client and number, with access restricted to people who administer that account. Number requirements are an onboarding input, not a reason to surround every purchase recommendation with a qualification exercise.
A monthly number-and-usage model
For a clean price comparison, use a separate US-only planning example: three clients, four local numbers each, 1,000 outbound minutes and 500 inbound minutes per client per month. Total: 12 numbers, 3,000 outbound minutes and 1,500 inbound minutes. Assume simple single-leg API calling without recording, SMS, transfers or AI. The UK estate above has a different tariff.
| Provider/basis | Monthly number rental | Outbound minutes | Inbound minutes | Telephony subtotal |
|---|---|---|---|---|
| Twilio Programmable Voice | 12 × $1.15 = $13.80 | 3,000 × $0.014 = $42.00 | 1,500 × $0.0085 = $12.75 | $68.55 |
| Plivo Voice API | 12 × $0.50 = $6.00 | 3,000 × $0.0115 = $34.50 | 1,500 × $0.0055 = $8.25 | $48.75 |
| Telnyx Voice API + starting US SIP rates | 12 × $1 = $12.00 | 3,000 × ($0.002 + $0.005) = $21.00 | 1,500 × ($0.002 + $0.0032) = $7.80 | From $40.80, excluding Managed Accounts commitment |
| Bandwidth US Voice API | Contractual rental for 12 numbers | 3,000 × $0.0100 = $30.00 | 1,500 × $0.0055 = $8.25 | $38.25 + number rental |
These are supported service bases, not all-in invoices. Telnyx's $40.80 is a pay-as-you-go usage illustration; it does not buy the managed-client account arrangement. Vonage can use the same formula with the applicable rate sheet. Dial managed calls would be 12 × $3 + 4,500 × $0.22 = $1,026 on its published basis, buying a different agent service. Self-hosted would be $621 before your own model costs.
Now include the agency's operating work. Assume a $20/month registry and monitoring allowance and two hours of administration at $75/hour. That adds $170, giving $238.55 for the Twilio example and $218.75 for Plivo. Neither amount includes initial application development, taxes or campaign labour.
Plivo saves $19.80/month on this traffic. A migration requiring an assumed eight hours at $75/hour costs $600, giving a simple payback of 30.3 months before disruption or duplicated service. For an established Twilio deployment, that is a weak migration case. For a new US-only application, Plivo's savings start immediately without a migration cost.
Number rental rarely dominates this modest-volume bill. Keeping six additional Twilio numbers for callbacks costs $6.90/month. Releasing them to save that amount can strand responses to a campaign the agency has already paid to run.
A forwarded or transferred call can introduce another billable leg. Recording, speech recognition and agent services can add further lines. Attribute those charges to the client and call, rather than dividing a blended provider bill equally across clients with very different traffic.
An inventory that survives client offboarding
Keep a small central registry that maps the provider number ID and E.164 number to the client, provider account, campaign, capability, routing destination and designated end-user organisation. Add the registration reference, intended retention date and offboarding destination. A nickname is helpful to humans; it should not be the application's unique identifier.
For example, client Northstar has two sales-campaign numbers and one enduring callback line. Its campaign ends on 31 October, but the callback line remains on proposal documents. The offboarding instruction should stop new outbound calls, route callbacks to Northstar's new destination and retain the number until the agreed transfer or port completes. It should not call a provider's release endpoint when the campaign record closes.
| Action | What changes | Appropriate use |
|---|---|---|
| Stop campaign activity | New outreach stops; number can stay active | End a campaign while preserving callbacks |
| Move between controlled accounts | Provider account assignment changes | Reorganise a number within a supported provider hierarchy |
| Port to another carrier | Carrier service changes; number is retained | Hand an enduring number to the client's chosen provider |
| Release/disconnect | The right to use the number ends | Dispose of a number after its dependencies have ended |
Twilio explicitly releases remaining numbers when a subaccount closes. Plivo's documented default on subaccount deletion transfers numbers to the main account unless cascade deletion is selected. Dial describes release as immediate and irreversible. Those differences make a universal “delete client account” automation inappropriate.
Use client-specific credentials for ordinary application work. Restrict parent-level provisioning, transfers and release to the agency's number service. Reconcile provider inventory against the registry so a number purchased manually does not become an unassigned expense or route to the wrong client. The operating principles in our signal-based outbound playbook apply here: explicit ownership, useful attribution and deliberate handover.
Which provider to choose
Start with Twilio for a new mixed-country agency application that needs documented client separation and a broad number lifecycle. Start with Plivo for straightforward US/Canadian client traffic where its porting scope is sufficient. Retain an incumbent provider when the savings are smaller than the work and risk of migration.
Choose Telnyx when aggregate spend or client-owned accounts make its account model practical. Choose Bandwidth when the inventory and communications product warrant a platform-level implementation. Keep Vonage where the existing API stack and subaccount access fit. Choose Dial for a suitable client-owned US agent project with a different application-cost proposition.
The strongest purchase is the provider whose account model and exit match the service the agency actually sells. The cheapest published minute cannot resolve a missing client boundary or a number that the client cannot retain.
FAQ
Does every campaign need its own subaccount?
Usually the client is the better account boundary. Several campaigns can share a client subaccount while keeping distinct number assignments and routing. Separate accounts per campaign make sense when permissions or contractual responsibility differ, but add administration and can make retained callback numbers harder to manage.
Does renting a number mean the client owns it?
Rental is a right to use a number under the provider's terms, not a property purchase. Record the client's end-user identity, who controls the provider account and who is authorised to request a port. An agency contract should describe the intended handover instead of treating possession of API credentials as ownership.
Can a number be moved without porting it?
Within a supported provider hierarchy, an account-assignment move may retain the carrier and number. Moving to another carrier is a port. Routing settings and messaging registrations may need separate work even when the number stays on the same provider. Release followed by repurchase is not a reliable way to retain it.
Why is Telnyx's $0.002 voice price not the complete call price?
It is the Voice API layer. The carrier/SIP leg is charged separately, producing a higher combined price before optional primitives. The client-management commitment is also a different commercial decision. Keep the API layer, carrier minutes and account arrangement visible in the budget.
When should the agency release a number?
After published references, active callbacks and client handover dependencies have ended. A quiet number can still appear in a proposal or old advertisement. Keeping a few low-rental numbers temporarily is often cheaper than losing an enquiry, especially when the original campaign cost is already sunk.
Is Dial an interchangeable alternative to Twilio for agency number management?
Its strongest role is a number attached to an AI-agent application. Current US provisioning, conditional porting and self-serve resale restrictions make it a different purchase. A client-owned account can fit that specialist use; a general agency estate should use a provider with the required client-account and migration model.
