PolyAI Pricing 2026: Every Cost Outside the Rate | Cekura

PolyAI pricing carries no published rate. Ongoing agent use is billed per minute, the rate is quoted per customer, and the number sits behind a demo form.

Budgeting for PolyAI pricing means reconstructing the bill from what PolyAI does put in writing, which turns out to be more than the pricing page suggests.

Everything below comes from PolyAI's own pricing page, product docs, and launch announcement, with every competitor rate checked against that vendor's published pricing.

TL;DR: PolyAI pricing

• PolyAI publishes no per-minute rate, no tiers, and no minimum commitment. The pricing page confirms the billing unit and lists what every plan includes.
• Since May 18, 2026, PolyAI has a self-serve path. Anyone can sign up at Agent Studio and build for free for the first two months. The company has not published what month three costs.
• The per-minute rate covers agent time. Telephony, most integrations, custom voices, and managed dashboards sit outside it, and several require an account manager.
• Wren, the natural-language builder inside Agent Studio, runs on a fair-use allowance PolyAI sets per account and does not publish.
• Skip the third-party dollar figures circulating for PolyAI contracts. None trace to a primary source, and this article does not repeat them.

PolyAI Pricing Plans: At a Glance

PolyAI ships no tiers. There are two ways in, plus a layer of costs that attaches to both. Here’s how the two paths compare:

PolyAI Pricing Plans Breakdown

Self-Serve Platform Access: Free for the First Two Months

What's included: On May 18, 2026, PolyAI opened its Agentic Dialog Platform to any builder with an email address, free for the first two months.

You get Poly Agent Builder for natural-language agent creation, the Agent Development Kit with self-serve API keys, and a shareable test environment for every agent. Raven runs as the default model, with GPT-5, Claude, and Gemini available as swaps.

The ADK installs with uv tool install polyai-adk and authenticates through poly start, which provisions an account and an API key in one pass. From there, the workflow is poly diff, poly push, poly chat.

Agent resources are YAML and Python files. You branch, review, and merge them like application code.

Best for: Engineering groups that want to hear Raven on a real call and read the resource model before anyone opens a procurement thread.

Pros:

✅ You reach a working agent without a sales conversation, which was impossible on this platform before May 2026.

✅ Quick Agent Setup crawls your public site and pre-populates topics and agent settings, so the starting point is closer than a blank project.

Cons:

❌ PolyAI has not published what happens at the end of month two. No rate, no conversion terms, no statement on whether a card is required at signup.

❌ Quick Agent Setup generates topics and settings only. Flows, variants, entities, handoffs, and integrations are yours to build.

❌ Self-serve accounts sit on the studio cluster. Enterprise workspaces live on us-1, euw-1, or uk-1, and poly start does not authenticate against them.

Enterprise Deployment: Custom Per-Minute Pricing

What's included: The pricing page confirms that ongoing voice agent use is billed per minute, and that the rate covers proactive performance improvements, maintenance, and 24/7 support.

Six things ship with every plan.

• A web ticket portal plus a 24/7/365 emergency support phone line
• Data infrastructure with compliance certificates and regular audits
• A 99.9% uptime SLA on your phone lines
• Monitoring and maintenance of system performance
• Upgrades to the latest platform features
• The underlying tech stack

Enterprise accounts are provisioned by PolyAI on a regional cluster. Sign-in runs through poly login --region us-1, and the ADK stores credentials per region so one machine can hold several.

Compliance coverage is deep. PolyAI holds ISO/IEC 27001 and SOC 2 Type II, operates to HIPAA and PCI-DSS requirements where relevant, complies with GDPR, and carries UK NCSC Cyber Essentials and Cyber Essentials Plus.

Best for: Financial services, healthcare, hospitality, and telecom operations where a caller reaching a dead line is a regulatory event.

Pros:

✅ One invoice covers the platform and the people who keep it running. Component-priced stacks leave you reconciling three vendors a month.

✅ Cyber Essentials Plus and ISO 27001 alongside SOC 2 Type II is unusually deep for this category, and it matters for UK public-sector procurement.

Cons:

❌ No published minimum, no published rate, no published rounding rule. You cannot model spend or compare quotes without entering the sales process.

❌ Enterprise workspaces cannot be self-provisioned. Only self-serve accounts can be created, so testing the actual deployment path requires the sales process first.

What Sits Outside the Per-Minute Rate

What this covers: This is the layer that decides whether a quote survives contact with production.

Telephony and numbers. PolyAI sells numbers directly in the US, UK, and Canada. Numbers elsewhere mean bringing your own Twilio account, which puts carrier charges on a separate invoice you negotiate yourself.

Managed-service integrations. A specific set of connectors is unavailable in self-service in Agent Studio. Custom SIP, DNI pooling, DesignMyNight, LiveRes (Zonal), Zendesk ticketing, PCI Pal, Stripe, Google Sheets, Ideal Postcode, and DeepL all start with an account manager.

Some of those require additional agreements or licensing with the third-party provider, which lands as a second contract to sign.

Custom voices. A brand-exclusive voice is available to enterprise accounts only and takes two to four weeks from recording brief to delivery. You supply the recordings and the legal clearance from the speaker.

Managed dashboards. PolyAI builds and maintains reporting dashboards for enterprise customers on Amazon Quick, formerly QuickSight. Changes route through your account manager unless QuickSight Authoring is enabled, which is itself gated to enterprise accounts and admin-permission users.

Wren usage. Wren is the natural-language builder inside Agent Studio, and it runs on a fair-use allowance. PolyAI states plainly that the size of the limit and its reset cadence are set per account and may change.

Each person gets their own allowance, resets land on UTC boundaries, and Wren API calls draw from the same pool. Once you hit that limit, the chat panel is replaced by a rest card carrying a Talk to Sales link.

Best for: Multi-country deployments, and any integration list that includes payments or a CCaaS handoff.

Pros:

✅ Bundled compliance work means fewer separate line items than platforms that price HIPAA as a monthly add-on.

✅ Wren's spend drivers are documented. Long chats, requests touching many flows at once, and repeated re-planning draw down fastest, so the allowance is manageable once you know the pattern.

Cons:

❌ An unpublished, adjustable per-account limit is a budgeting problem. You cannot tell whether your team will hit the ceiling in week one or never, and PolyAI can change the number without notice.

❌ Numbers outside three countries mean a second vendor relationship and a second invoice from day one.

❌ Custom voice and managed dashboards both route through a human at PolyAI, which adds calendar time to any change.

Which PolyAI Plan Should You Choose?

Choose self-serve if you:

• Want to evaluate Raven on live audio before committing engineering time
• Build in a Git-backed workflow and want to read the resource model directly
• Have a low-stakes internal use case that fits inside two months of free access

Choose enterprise if you:

• Run regulated phone traffic where a 99.9% SLA is a contractual requirement
• Need Custom SIP, PCI Pal, or a CCaaS handoff, all of which require an account manager
• Operate in a region with data residency requirements the studio cluster does not satisfy

Choose a platform with published rates if you:

• Need a per-minute number before a budget cycle closes
• Run under 10,000 minutes a month
• Want to model three configurations and compare them without a call

Take five questions into the quote conversation, because the answers are what separate a quote from a forecast.

• Rounding rule: Is a 41-second call billed as 41 seconds or as a full minute? At 50,000 calls a month, per-minute rounding on a 41-second call adds roughly 16,000 billed minutes you never used.
• Billable definition: Does the meter run during hold, during silence, and after a transfer to a human?
• Minimum commitment: What annual volume does the quoted rate assume, and what happens if you land under it?
• Wren allowance: What size and cadence is set on your account, and is it negotiable in the contract?
• Separate invoices: Which integrations carry third-party licensing, and who signs those agreements?

Is PolyAI Worth the Cost?

Whether PolyAI pricing is worth it depends on whether you are buying a platform or buying an outcome.

PolyAI sells the outcome. The per-minute rate absorbs maintenance, monitoring, upgrades, and 24/7 support, and the compliance stack arrives without a line item.

For a contact center with no in-house conversational AI engineering, that bundle is worth a premium, because the alternative is hiring for it. For engineering-led organizations already running a voice stack, the same bundle is work you have already staffed.

A worked example, illustrative only. Take 40,000 inbound calls a month at an average handle time of 2.5 minutes. That is 100,000 agent minutes.

At a hypothetical $0.30 per minute, the platform line reads $30,000 a month. Add US telephony at $0.015 per minute, Retell's published Twilio-routed rate, and it moves to $31,500.

These figures are a modeling exercise, and PolyAI publishes no rate against which to check them.

Cut average handle time to 2.0 minutes and the same call volume bills 80,000 minutes, or $24,000 at the same hypothetical rate. That $6,000 swing comes entirely from handle time.

Run that model with your own handle time first. Handle time moves the bill more than the rate does, and it is the one input you already know. Our guide to voice agent pricing models walks through how each structure behaves at volume.

PolyAI is worth it if you:

• Handle high-complexity calls where turn-taking and interruption recovery decide the outcome
• Need ISO 27001, SOC 2 Type II, and Cyber Essentials Plus under one contract
• Have no conversational AI engineering function and no plan to build one

Skip PolyAI if you:

• Need a published rate to close a budget cycle
• Run modest volume where an annual enterprise agreement adds procurement cost for no gain
• Want to assemble your own speech and model stack, since PolyAI's value sits in the parts it manages for you

PolyAI Alternatives & Pricing Comparison

Three platforms in this category publish component rates. Every figure below was verified against each vendor's own pricing page on 27 August 2026. Verify with the vendor before you commit.

Managed enterprise CX vendors like Cognigy and Parloa sit in PolyAI's actual competitive lane, and none of them publish a rate card either, which is why the comparison above uses platforms that do.

You'll find the contrast sharper on the details than on the headline numbers.

Retell bills to the nearest second with no per-call rounding, and says so on its pricing page. Our breakdown of what a Retell minute really costs shows how the components stack.

Vapi prices HIPAA as a $2,000 monthly add-on and Zero Data Retention at $1,000, so the compliance premium is a number you can put in a spreadsheet.

ElevenLabs caps concurrency by tier, from 4 concurrent calls on the entry plan and 6 on Starter, up to 40 on Business.

Per Cekura's benchmarks, which ran 82 scenarios across seven voice agent configurations, reliability across three consecutive runs ranged from 75.6% on Retell down to 30.5% on Gemini Live, and mean response time from 1.27s on ElevenLabs to 3.08s on Vapi.

PolyAI answers none of those three questions in public. That is the practical difference between PolyAI and the platforms above: you cannot model the spend before you enter the sales process.

When Sales-Led Pricing Works in Your Favor

Volume commitments buy negotiating room. At scale, an all-in rate covering support, monitoring, and upgrades often lands below a component-priced stack once you cost the engineering hours those components require.

Bundled compliance is the clearest case. Vapi's HIPAA add-on alone runs $24,000 a year before a single minute of traffic, and that sits on top of the per-minute rate rather than inside it.

When a Published Rate Serves You Better

Below roughly 10,000 minutes a month, the procurement cycle costs more than the contract saves you.

Published component rates also let you model the lever that matters most, which is model choice. On Retell's published rate card, swapping GPT-5.5 at $0.16 per minute for Claude 4.5 Haiku at $0.025 cuts the model line by 84% with no contract renegotiation.

Cekura vs PolyAI: Which Should You Choose?

PolyAI runs the agent, and Cekura tests it, including agents running on someone else's platform.

PolyAI ships simulation tests, A/B testing, and CI gating through the ADK CLI and API. Those work well, and they work on PolyAI agents. Cekura tests whatever you ship on, which matters the moment your stack includes more than one platform, or you are still deciding between them.

PolyAI is better for: Running the agent itself. Raven, the speech pipeline, barge-in handling, and the managed deployment are the product, and platform-native testing is included in the rate you already pay.

Cekura is better for: Independent verification across platforms. Cekura runs pre-production simulations and monitors live conversations on top of Retell, Vapi, ElevenLabs, LiveKit, Pipecat, Bland, and more, so a bake-off between vendors uses one scoring harness on both sides.

Use both if: You are evaluating PolyAI against another platform, or running PolyAI alongside a second stack. A vendor's own test suite cannot tell you how that vendor compares to the alternative.

Where Cekura fits across the lifecycle:

• Pre-production: Simulate hundreds of caller scenarios with custom personas and accents before a number goes live, then red-team the agent against jailbreaks and adversarial callers.
• Infrastructure: Test how the agent handles interruptions, background noise, and latency under concurrent load, which is where per-minute costs and containment rates both move.
• Observability: Monitor live conversations for instruction-following, tool-call accuracy, and CSAT, with Slack alerts on latency spikes and quality drops.

Cekura is SOC 2, HIPAA, and GDPR compliant, with transcript redaction, role-based access, and full audit trails.

Comparing PolyAI against a platform with published rates? Book a demo and score both on the same scenarios before the contract conversation starts.

The Bottom Line on PolyAI Pricing

PolyAI pricing rewards buyers who want a vendor to own the outcome, and penalizes anyone who needs a number this quarter.

The platform is genuinely strong on hard calls, the compliance stack runs deeper than most of the category, and the bundled support is real value for an operation with no conversational AI engineers.

The self-serve opening in May 2026 changed the evaluation, though it did not change the purchase. You can now hear the product before you talk to anyone, which is the right way to start.

Model your own handle time, take the five quote questions into the call, and get the rounding rule in writing.

Frequently Asked Questions

How Much Does PolyAI Cost?

PolyAI costs a per-minute rate the company does not publish. The pricing page confirms the billing unit and states that the rate includes maintenance, upgrades, and 24/7 support, then routes buyers to a demo form. Dollar estimates on third-party sites trace back to no primary source.

Does PolyAI Have a Free Trial?

Yes, PolyAI has offered free self-serve access since May 18, 2026, when it opened the Agentic Dialog Platform to any builder free for the first two months. PolyAI has not published what the platform costs after that period ends.

Does PolyAI Publish Per-Minute Rates?

No, PolyAI does not publish per-minute rates. The pricing page states the billing unit and lists six inclusions that ship with every plan, including a 99.9% uptime SLA, without naming a figure or a minimum commitment.

What Is the Difference Between PolyAI Self-Serve and Enterprise?

The main difference between PolyAI self-serve and enterprise is where your workspace runs and what you can configure. Self-serve accounts sit on the studio cluster. Enterprise workspaces run on clusters like us-1, and they open up custom voices, dashboards, and gated integrations.

Is PolyAI HIPAA Compliant?

PolyAI states its systems are designed to meet HIPAA requirements where relevant and handles protected health information under those controls, though it does not publish a HIPAA certification the way it does ISO/IEC 27001 and SOC 2 Type II.

What Is Not Included in the PolyAI Per-Minute Rate?

Telephony outside the US, UK, and Canada sits outside the PolyAI per-minute rate, along with managed-service integrations such as Custom SIP and PCI Pal, custom brand voices, and managed dashboards. Wren usage also draws on a separate per-account allowance.