WebRTC softphone, inbound routing, IVR, voicemail and stereo recording. Telephony at the core, other channels as a blend.
All featuresTranscription, checking against your checkpoints, and red flags. Every verdict comes with the citation it rests on.
All featuresOwn checkpoints per workspace. Energy on the Quality Standard, insurance on conduct rules, you on your own brand rules.
All use casesCallCheckPro checks before every outbound attempt whether you are allowed to call this person, then reviews the call against your checkpoints. The legal basis, the recording, the verdict and the citation it rests on all sit with the same call.
Three operations call with it every day. Bwiser also runs the platform itself, so whatever fails on the floor reaches us the same day.
One call touches three systems: the calling itself, the review afterwards, and the steering that follows. CallCheckPro does all three, on the same data.
Outbound dialer or inbound routing
Transcription + 33 checkpoints + red flags
Live dashboards, trend reports, coaching
Since 1 July 2026, Dutch law no longer only asks whether a call went well, but whether you were allowed to make it. CallCheckPro answers that question before the line opens — inside the call route itself. Not in a screen an agent can click past.
Global, per calling organisation, or only for this campaign. When someone withdraws consent or exercises their right to object, the number lands here automatically.
An unsolicited commercial offer falls under the consent requirement; service and performance of an existing contract do not. If the purpose is left blank, the gate closes rather than opens.
A limited company you may call. For a consumer, sole trader or partnership, only a recorded consent file that is valid at the moment of calling counts. If the legal form is unknown, the platform looks it up at the Chamber of Commerce instead of guessing it from the company name.
A minimum interval and a cap per number per calling organisation. Leads that are not clear yet are simply not handed out, so in progressive dialing the agent never even sees them.
The outbound number is tracked for reachability and behaviour. If one drops off, the gate rotates to a number that still works.
Set per workspace, and overridable per campaign with a reason attached.
Measuring without intervening. Every call gets a verdict in the call log, but nothing is stopped. That way you first see how much a hard gate would block before you switch it on. This is where a new workspace starts.
The call runs, but the lead is flagged as called without a legal basis and the exception goes into the audit trail. Useful while your consent records are not yet complete and you want to see who it concerns.
No legal basis, no call. In this setting a merely plausible consent no longer counts: only an opt-in the server measured itself gets through. A one-off release stays possible, traceable to the person who granted it and the reason they gave.
Consent you collect yourself, along a route you can show afterwards.
A hosted form per channel, or a connection to your own site through the API. An intake is not yet consent — the gate does not read it.
The person confirms themselves: a click in an SMS, a double opt-in email, or a call to a verification number. Only that step, measured by the server, produces certainty proven.
A SHA-256 hash covers the evidence fields, after which they are fixed. A verification link lets an opt-in be checked from outside the platform too.
Records from an import stay plausible — which is honest, because nobody measured that confirmation. A re-confirmation campaign sends them up the same route.
What you see below is the running application, not an animation. You follow one call through the platform: the gate before it, the recording, the review, and the report the supervisor has in front of them the next morning. 95 seconds.
A supervisor listens back to a handful of calls a week. The rest goes unheard, so coaching is about whatever someone happened to catch.
The agent works the call and notices nothing of what happens underneath: it is recorded in stereo — client and agent on their own channel — stored encrypted and queued for review.
A verdict per checkpoint with the transcript citation and the timestamp, so you can listen back to it. If the agent disagrees, they file an appeal and a person decides.
Which checkpoints keep failing, for whom, and is it moving the right way after a training? That sits in the trend report, not in someone's memory.
From the softphone in the browser to the admin screen where you set up campaigns, queues, checkpoints and retention periods. Twenty-eight admin sections, without a line of code.
Every deal is reviewed, plus a sample per agent drawn across the day. If you want to move towards 100% of your calls, that is a question of GPU capacity — we work through your volume up front.
Every line carries a hash of the previous one, so a deleted or altered line stands out. Demonstrably regulator-compliant.
Call from the browser. No hardware. No VoIP app. Works on desktop and mobile.
Own servers in the EU. No cloud-LLM tokens.
Live overview of all agents, KPIs, queues and compliance. Coach on patterns, not on samples.
Tone and escalation risk, so you pick out the sensitive calls first.
Cooldowns, max attempts, intelligent routing.
Separate workspaces per client.
Every AI report is contestable. A human decides. No black box.
Alongside the checkpoints, an assessment of the risk of a complaint or cancellation, with concrete improvement points drawn from the call itself.
CallCheckPro runs on its own; you need no other system alongside it. RelayPro is natively connected — click-to-call, call history and outcomes flow straight through. For everything else there is a REST API over a WireGuard tunnel, and we build the integration together with your vendor.
RelayPro is the only integration that ships ready to use. The other systems connect through the REST API; which fields move back and forth is settled during setup.
Calling with another system but still want the review? Push your audio over a WireGuard tunnel to the API. The endpoint is not on the open internet — outside the tunnel there is nothing to knock on.
# 1. Activate WireGuard tunnel
wg-quick up callcheckpro
# 2. Push audio
curl -X POST https://10.0.0.1/v1/analyse \
-H "Authorization: Bearer $API_KEY" \
-F "audio=@call.wav" \
-F "agent_id=mv-012" \
-F "campaign=energy-q2"
# Response
{
"analyse_id": "a-2847-xk9",
"score": 91,
"checkpoints_passed": 30,
"checkpoints_total": 33,
"red_flags": 1,
"status": "complete"
}The 33 checkpoints you hear us talk about come from the energy sector, because that is where the platform grew up. The engine underneath simply reads the criteria you set — per campaign, per workspace.
Four categories of checkpoints covering opening, offer, terms and conduct — from "which supplier are you calling for" to the fourteen-day cooling-off period. Plus eleven red-flag signals, including calling without the explicit opt-in that has been mandatory in the Netherlands since 1 July 2026.
Progressive dialing with a pre-caller/closer split, and checkpoints on your pitch, your objection handling and the follow-up appointment. Coaching on what was actually said, with the citation attached.
Queues with priority, call menus, opening hours, voicemail and a one-key call-back request from the queue. Every call is recorded. The AI review is currently tuned to cold outbound; the inbound variant of the checkpoint set is on the roadmap.
Offers explained correctly, no promises that cannot be kept, and records that hold up. Exactly the calls where a complaint costs you most.
Duty of care, disclosure obligations, correct product explanation. Configure checkpoints based on regulatory requirements or internal quality standards.
Telecom, healthcare, government, e-commerce. You write down what a good call has to contain in your business; we turn that into a checkpoint set and run it.
Not logos of companies that once sat through a demo. These are the three operations that call with it every day, and the people who run them.
Energy sales for Eneco. Bwiser also runs CallCheckPro itself, so whatever fails on this floor comes back to us before another client notices.
Sales and service in one operation: outbound campaigns alongside the inbound line, in the same workspace.
Sales and service, with AI review of their sales calls: their own checkpoints rather than the energy preset.
Which is why not a single recording sits with somebody else's service. Transcription and review run on our own GPU in a European data centre, and the recording itself sits encrypted on our own disks.
Built to the requirements of these standards — not held as separate certifications. We share our security documentation on request under NDA.
Dedicated servers in a European data center. No shared hosting. No third-party multi-tenant cloud.
Transcription and analysis on own GPU servers. No external traffic with customer calls. No cloud-LLM tokens.
VPN tunnel, mutual TLS with client certificates, API-key auth and cryptographic request signing.
Data processing agreement as standard. Configurable retention periods. No processing outside the EU. DPIA on request.
Chain-linked logging: every line contains a hash of the previous one. Tampering is detectable. Directly exportable.
Role-based access with team scoping, and sensitive rights per person: who may download recordings is enabled per supervisor. Two-factor authentication, and rate limits per IP and per user on the heavy endpoints.
Start with one team and scale when you need to; an annual contract is not required. Call rates sit separately from the licence, so you can see where your money goes.
Each license includes 2,500 call minutes/month. Beyond that the rates below apply. Billed per second, no setup fee.
CallCheckPro is a calling platform with a compliance layer around it. Your agents call out of campaigns through a softphone in the browser, or take incoming calls from a queue. Before every outbound attempt the platform checks whether you are allowed to call this person; afterwards the call is recorded and reviewed against the checkpoints you configured. Supervisors steer on live dashboards, administrators set everything up across 28 admin sections. It all runs on our own servers in the EU.
Every deal is reviewed: every call with an outcome that counts as a success. On top of that the platform draws a sample per agent, spread across the day so it is not all morning calls. If you want to move towards reviewing all calls, that is mainly a capacity question — the review runs on our own GPU and takes roughly a minute per call. We work through your volume up front and add hardware where needed. Voicemails and calls shorter than the configured minimum duration are skipped either way.
Yes. Outbound with progressive dialing, campaign management and lead assignment. Inbound with queues, routing strategies (longest-idle, round-robin, skills-based), call menus, opening hours, voicemail and a one-key call-back request from the queue. You can set up workspaces that combine both. The AI review is currently tuned to cold outbound; an inbound variant of the checkpoint set is on the roadmap.
Every call is recorded in stereo — agent and client on their own channel — stored encrypted and transcribed on our own GPU. A local language model then walks through your checkpoints and gives a verdict per checkpoint with the transcript citation it rests on. Separately from the score there are eleven red-flag signals, and you get a risk assessment with concrete improvement points. No audio and no transcript leaves our network.
For legal entities nothing changes; you may still approach them with an unsolicited commercial offer. For consumers, sole traders and partnerships explicit consent has been required in the Netherlands since 1 July 2026: the existing-customer exemption has lapsed. CallCheckPro assesses this per attempt — it determines the legal form, looks it up at the Chamber of Commerce where needed, pulls in the consent record and stores the verdict with the call. Whether that verdict also stops the call is up to you, through the gate setting. This is not legal advice; we supply the record, you keep the judgement call.
Through the opt-in portal. You publish a hosted form per channel or connect your own website through the API. What comes in is not yet consent: the person has to confirm themselves, with a click in an SMS, a double opt-in email, or a call to a verification number. Only that step, which the server measures itself, produces a record with certainty 'proven'. A SHA-256 hash covers the evidence fields, after which they can no longer be changed, and a verification link lets an opt-in be checked from outside the platform. Imported consents stay 'plausible' until a re-confirmation campaign has taken them up the same route.
All of them. For energy telemarketing we ship a preset of 33 checkpoints across opening, offer, terms and conduct, each with its own weight. Beyond that you write your own set: greeting, tone of voice, upsell policy, complaint handling, correct disclosure. Prompts and scoring weights are versioned in the admin screen, and you can activate a different set per campaign or per client.
Then you appeal. The agent or supervisor backs it up with counter-evidence from the transcript, an administrator reviews it and adjusts the score if the appeal is justified. The original status is preserved, and both the appeal and the decision go into the audit trail. Without that button no agent accepts coaching from a model, and rightly so.
Everything runs on our own servers in an EU data centre, and transcription and review happen locally; no calls go to external services. Every workspace automatically receives a data processing agreement at onboarding. Retention periods are set per workspace, and we provide a DPIA on request. Recordings are stored encrypted, and every time someone plays one back or downloads it, that is logged.
RelayPro is natively connected: click-to-call, call history and outcome registration flow through without setup. Other systems connect through the REST API, which sits behind a WireGuard tunnel and is therefore not on the open internet. Which fields move back and forth is settled during setup, together with your vendor.
You set that per workspace. Recent recordings sit on fast storage so you can play them back instantly; after that they move automatically to encrypted archive storage overnight, with a floor of 91 days. The archive is sample-checked daily against the stored hash, and once a month we pull a blob back and decrypt it to prove the recovery path works. Transcripts are wiped after a year; summary, checkpoints, score and hash remain.
Starter from €49 per agent per month, Professional from €89 with the full review and reporting, Enterprise on request. Every licence includes 2,500 call minutes per month to Dutch landline and mobile numbers. Beyond that: NL landline 1.9 ct/min, NL mobile 6.0 ct/min (excl. VAT), billed per second with no setup fee.
No sales team and no funnel: you talk to the people who built it. Send a recording of one of your own calls and in twenty minutes you will see what the review pulls out of it, and where your checkpoints should sit differently.
+31 (0) 85 130 7428