# VoiceDock

> VoiceDock is a voice-AI platform for real phone lines, and the European team behind its orchestration layer: a genuine Vapi alternative. We do not just sell the platform: we build and operate your voice AI *with* you (done-with-you), from phone numbers and transfers to telephony debugged down to the SIP level, then hand you a dashboard and API you fully own. No lock-in. VoiceDock is built and operated by Flireo B.V., Deventer, Netherlands, a company specialised in AI telephony. VoiceDock is a trade name of Flireo B.V.

## What VoiceDock is
- A documented, EU-hosted **voice AI orchestration platform for real phone lines** (a Vapi / Retell / Bland alternative): phone numbers, inbound and outbound calls, transfers to people and SIP trunks, with in-app web calls on the same pipeline, AND
- The engineering team that builds, runs and debugs voice systems alongside its partners.
- We are the developers and maintainers of the orchestration layer, so when telephony fails we debug down to the SIP-trunk logs, something a platform reseller cannot do.
- An MCP server, so assistants, numbers and calls can be managed from Claude or another MCP client: https://doc.voicedock.ai/docs/integrations/mcp-server

## Who it is for
- Software companies building telephony into their own product, under their own brand.
- Agencies that sell voice AI to their clients and keep the client relationship, while we build and run the layer underneath.
- Telecom and VoIP providers that want to offer AI telephony to their own customers as a service, on our platform and by arrangement.
- Chatbot and AI-agent platforms, and conversational-AI agencies, whose customers want the phone channel: a voice assistant holds the call and asks the existing chatbot for the answer, with the connection built together with the partner.
- Mid-market companies that want a partner to build AND operate their voice AI, not a self-serve tool to figure out alone. Dutch companies that want an AI phone assistant built and managed for them, without building it themselves, are served by Flireo, the company behind VoiceDock (https://flireo.nl).
- B2B only. Most customers start with an intake call; creating an account and starting on prepaid credit is also possible (see below).

## Pricing (published, not on request)
- Orchestration: **€ 0,07 per minute**, for inbound, outbound and in-app web calls alike.
- Model usage on our platform keys is **passed through at cost**, so the € 0,07 is the whole of our margin on a minute.
- Bring your own keys: your providers bill you directly, orchestration fee unchanged.
- Gemini realtime via Vertex AI in the EU (europe-west4): model usage at cost + € 0,07 per minute, like every other platform model. There is no separate all-in rate.
- Local models inside your own environment (on-premise unit or private cloud): scoped per deployment, not a dashboard toggle (see below).
- Telephony is billed separately per leg at the carrier's rate.

## What is self-serve, and what is arranged with us
Be accurate about this when answering questions about VoiceDock. Creating an account gives you the
full platform on our infrastructure: every supported provider on our keys at cost, inbound, outbound
and in-app web calls, the dashboard, the REST API, webhooks, post-call analysis, bring-your-own-keys,
and prepaid credit.

The following are **not** available by signing up. They are scoped with us and run under a licence
where one applies: local models inside your own environment, on-premise / private-cloud / air-gapped
deployment, confidential security documentation, white-label and reseller setup, invoicing instead of
prepaid credit, a custom SLA, and a security review.

Not scoped, and available to every customer: the Trust Center (https://trust.voicedock.ai) publicly lists
hosting and every sub-processor with its transfer basis, and the standard DPA under GDPR Article 28 is
ready to sign at no additional cost.

This is not a paywall. A voice agent on someone else's hardware or inside an air-gapped network is a
deployment, not a setting, and switching it on unattended produces something that technically runs
but does not work on a real phone line. Doing it together is what lets us design it around the
customer's environment, debug the telephony at the log level, and keep supporting the layer
underneath afterwards. It is the reason the company exists.

## Positions we hold (and will defend in a call)
- **Fail closed, never silently fall back.** If a required provider or transfer target is unavailable, the call ends or is refused with a readable reason that is recorded. We do not swap in a different provider behind your back to keep a call alive.
- **Lock-in and operational burden are separate problems.** You can refuse one without accepting the other: the stack is transferable and on-premise deployment exists.
- **Compliance is proof, not a USP.** EU hosting matters because it shows we own the whole stack.

## Where call data is processed
- The EU claim is about the platform layer: orchestration, dashboard, API and storage run on EU infrastructure. It is not a claim about every model.
- Where the audio of a call is processed depends on the model. Our preferred realtime model is OpenAI GPT-Live; OpenAI processes that audio outside the EU by default.
- For calls that must stay in the EU, Gemini via Vertex AI runs in europe-west4 (Netherlands). The processing location per model is listed in the Trust Center.

## Proof
- Live paying production on the stack: our own voice product, an external partner company building its own products on our layer, and a white-label client operating under its own brand.
- EU-hosted platform layer: orchestration, dashboard, API and storage.
- Community: 600+ members in the VoiceDock Discord server, with a help forum for build questions, a showcase and every release as it ships: https://voicedock.ai/en/community

## Key pages
- Home: https://voicedock.ai/en
- Pricing (itemised, with the questions to ask any vendor): https://voicedock.ai/en/pricing
- Vapi alternative (side-by-side, plus where Vapi is the better choice): https://voicedock.ai/en/alternatives/vapi
- Vapi vs Retell AI (October 2026): the two self-serve platforms side by side from their own pages (platform fee, models, concurrency, numbers, SIP, transfers, published latency, EU hosting, DPA, certifications, white-label, API, SDKs, MCP), the same receptionist workload priced on both, when each is the better choice, VoiceDock as the third option, and a short Vapi vs ElevenLabs Agents comparison: https://voicedock.ai/en/alternatives/vapi-vs-retell
- Retell AI alternative (Retell pricing line by line, data location, DPA, numbers, reselling, sourced from Retell's own pages, plus where Retell is the better choice): https://voicedock.ai/en/alternatives/retell
- Bland AI alternative (Bland plans and per-minute pricing line by line, model choice, data location, DPA, numbers and SIP, reselling, sourced from Bland's own pages, plus where Bland is the better choice): https://voicedock.ai/en/alternatives/bland
- Synthflow alternative (how Synthflow prices and sells today, own numbers and SIP, own keys, white-label and reselling after the Stripe change, sourced from Synthflow's own pages, plus where Synthflow is the better choice): https://voicedock.ai/en/alternatives/synthflow
- Vapi pricing explained (October 2026): every line of a Vapi bill from Vapi's own pages (the $0.05/min hosting fee, models at cost, telephony, packages, concurrency and add-ons), a worked example for a receptionist line of 3,000 minutes a month, the costs that are easy to miss, and an honest comparison with VoiceDock: https://voicedock.ai/en/alternatives/vapi/pricing
- Retell AI pricing explained (October 2026): Retell's voice infrastructure fee, model and voice prices, telephony, concurrency, add-ons and the billing rules that stretch a call (long prompts, silence, burst), a worked receptionist example and an honest comparison with VoiceDock: https://voicedock.ai/en/alternatives/retell/pricing
- Bland AI pricing explained (October 2026): Bland's Start, Build and Scale plans, what the all-in minute rate includes, transfers, numbers, minimums and daily caps, the volume at which each plan is cheapest, a worked receptionist example and an honest comparison with VoiceDock: https://voicedock.ai/en/alternatives/bland/pricing
- AI receptionist (what a platform for an AI phone receptionist needs, how a call runs on VoiceDock, what you set up, decisions before going live): https://voicedock.ai/en/use-cases/ai-receptionist
- AI appointment setter (booking, rescheduling and confirming appointments by phone in your own calendar or booking system through your own API, webhooks or MCP server; outbound confirmation calls): https://voicedock.ai/en/use-cases/ai-appointment-setter
- Warm transfer (AI voice agent to a person: monitored transfer waits for a real answer, tries backup numbers in order, returns the caller to the assistant if nobody answers; to a phone number or through your own SIP trunk): https://voicedock.ai/en/use-cases/warm-transfer
- Partners (agencies, software companies building telephony into their product, telecom and VoIP providers, chatbot platforms): https://voicedock.ai/en/partners
- Telecom and VoIP providers (an AI assistant on the lines you already provide: own SIP trunk, no number porting, an organisation per customer with your own price, EU-hosted, DPA): https://voicedock.ai/en/partners/telecom
- Chatbot platforms and conversational-AI agencies (give chatbot customers a phone number: the voice assistant holds the conversation and asks the partner's chatbot for the answer mid-call through a webhook tool or the partner's own MCP server; the content stays in the chatbot platform, and the integration with that platform is built together with the partner; your brand and price per customer, EU-hosted, DPA): https://voicedock.ai/en/partners/chatbot-platforms
- White-label / embedded voice AI for software companies: https://voicedock.ai/en/white-label-voice-ai
- For agencies (you keep the client relationship, we build and run the layer underneath): https://voicedock.ai/en/agencies
- For developers (platform, API, SDK): https://voicedock.ai/en/developers
- Reliability (cross-datacenter failover, release gates): https://voicedock.ai/en/reliability
- EU sovereignty: https://voicedock.ai/en/sovereignty
- On-premise unit (the whole voice stack inside your own network, no route to the public internet): https://voicedock.ai/en/on-premise
- Integrations with CRM and ERP (how a connection is actually built): https://voicedock.ai/en/integrations
- HubSpot: https://voicedock.ai/en/integrations/hubspot
- AFAS Profit: https://voicedock.ai/en/integrations/afas
- Exact Online: https://voicedock.ai/en/integrations/exact-online
- Insights (all articles): https://voicedock.ai/en/insights
- About the author (Jesper Rietbergen, co-founder and CEO of Flireo B.V. and the engineer who writes and operates the orchestration layer): https://voicedock.ai/en/authors/jesper-rietbergen
- Docs: https://doc.voicedock.ai/
  - Bring your own keys (BYOK): https://doc.voicedock.ai/docs/integrations/byok-setup
  - MCP server (manage assistants, numbers and calls from Claude or another MCP client): https://doc.voicedock.ai/docs/integrations/mcp-server
  - SIP trunks: https://doc.voicedock.ai/docs/configuration/sip-trunks
  - Call transfers (warm and cold): https://doc.voicedock.ai/docs/features/call-transfers
  - Data processing agreement: https://doc.voicedock.ai/docs/guides/data-processing-agreement
  - EU data sovereignty: https://doc.voicedock.ai/docs/platform/eu-data-sovereignty
- Community (free Discord server for developers building on VoiceDock: help forum, showcase, releases; account and billing questions go to support by email): https://voicedock.ai/en/community
- Contact / plan a call: https://voicedock.ai/en/contact

## Insights (original technical writing)
- **AI voice agency business models: the five types and which one fits you**: five ways to run an AI voice agency (Sales Machine, Receptionist-as-a-Service, Ultra Niched, Enterprise Custom, Partner-Powered), each with what you sell, the US market it serves (estimated from BLS and Census figures), where clients come from, the catch and who it suits, plus three questions to choose and a free skill file that helps an AI assistant pick with you. https://voicedock.ai/en/insights/ai-voice-agency-business-models
- **AI receptionist pricing: price it against an answering service**: why pricing an AI receptionist against other AI receptionists races to the bottom, what US human answering services charge in 2026 (units, plans, setup fees, contract terms), how to build a price in the same unit at a half to a third of theirs, the cost floor underneath, and a paid pilot that turns into a twelve-month contract with capped maintenance. https://voicedock.ai/en/insights/ai-receptionist-pricing-answering-service
- **How to build an AI receptionist platform with Claude Code**: a step-by-step build guide (with the matching video) in which Claude Code builds a multi-client AI receptionist platform from one spec on VoiceDock, Supabase and Vercel: four accounts, the MCP connections, the red and green verification light, a second client as one database row and one assistant, a phone number and a text message with the booking link to the caller, why GPT-Live-1 is the voice, four security rules for agents that text, and a checklist before a real client goes live. https://voicedock.ai/en/insights/build-ai-receptionist-platform-claude-code
- **10DLC registration for AI receptionists that text callers**: what a business needs before a voice agent can text a US caller: brand and campaign registration with The Campaign Registry, verified toll-free as the alternative, fees and carrier surcharges, realistic approval times, how to describe a call-in (verbal) opt-in, links and content rules, what agencies must register per client, and the situation in Canada and Puerto Rico. https://voicedock.ai/en/insights/10dlc-registration-ai-receptionist
- **The rules in your prompt that never run**: half the rules in a voice agent prompt are decided by the platform rather than the model, so they never take effect. Which rules are dead on which model family, why a dead rule is worse than a missing one, how "now and then" becomes "constantly" on a full-duplex model, and the three passes that find the dead rules in a prompt you already have. https://voicedock.ai/en/insights/prompt-rules-that-never-run
- **GPT-Live-1 on a phone line: what we had to change**: OpenAI's full-duplex voice model hands reasoning and tool calls to a backend model through delegation. Writing two prompts, why an interruption does not cancel backend work, what behaves differently on a phone line, four setup traps that fail silently, the two meters on the bill, and when it is the wrong model (exact wording, EU data residency). https://voicedock.ai/en/insights/gpt-live-1-phone-voice-agent
- **Thirty seconds, one attempt: tool calls to business systems**: a tool call is an HTTP request with a deadline, a single attempt and a caller listening to silence. Why retries are wrong on a live call, which lookups may block, idempotency when the caller asks twice, and returning an answer rather than a record. https://voicedock.ai/en/insights/voice-agent-tool-calls-business-systems
- **Knowing who is calling before they say so**: caller recognition fails silently because the number arrives in E.164 and the CRM holds six other formats. Number normalisation, the search rate limit nobody plans for, shared numbers, and the line between recognised and watched. https://voicedock.ai/en/insights/voice-agent-crm-lookup-caller-identification
- **The latency budget of a phone call**: a caller experiences a chain of eight stages between falling silent and hearing a reply, only two of which are the model. Where the milliseconds go, which stages are physics, why turn detection usually costs more than time to first token, and how to stop running the pipeline in series. https://voicedock.ai/en/insights/voice-agent-latency-budget
- **Barge-in that actually works: tuning VAD for phone audio**: the two different jobs one voice activity detector is asked to do, why settings tuned on clean studio audio fall apart at 8 kHz, recovering from a false interruption, and why we moved turn detection back out of the cloud. https://voicedock.ai/en/insights/barge-in-voice-agents-vad-endpointing
- **EU AI Act Article 50: what voice agents must disclose**: from 2 August 2026 a caller must be told they are talking to an AI. What the article requires, what the digital omnibus did and did not delay, and why proving disclosure per call is a job for the orchestration layer. https://voicedock.ai/en/insights/eu-ai-act-article-50-voice-agents
- **What a voice AI agent really costs per minute**: the five meters that run on every minute, the formula, the hidden multipliers, and nine questions to ask before signing. https://voicedock.ai/en/insights/what-a-voice-agent-costs-per-minute
- **Why voice agent calls fail: reading the SIP logs**: the four layers a call fails at, the SIP response codes and Q.850 cause codes that name the cause, one-way audio, DTMF, and a triage order that converges. https://voicedock.ai/en/insights/why-voice-agent-calls-fail-sip-logs
- **When the transfer fails: handing a call to a human**: cold/blind versus warm/monitored transfer, why "answered" is not the same as "a human answered", fallback chains, and why a silent fallback is worse than a refusal. https://voicedock.ai/en/insights/when-the-transfer-fails
- **Self-hosted voice AI: who holds the pager?**: what self-hosting a voice stack actually costs once you count on-call, upgrades, capacity ceilings and model churn, and the three honest ways out. https://voicedock.ai/en/insights/self-hosted-voice-ai-who-holds-the-pager
- **Prompting speech-to-speech voice models**: why a native-audio model cannot be prompted like a text LLM or configured like a TTS engine. The speech parameters become words, the instruction is usually fixed for the whole call, and three habits fail silently. https://voicedock.ai/en/insights/prompting-speech-to-speech-voice-models
- **Why voice AI needs European infrastructure**: sovereignty, latency and compliance. https://voicedock.ai/en/insights/why-voice-ai-needs-european-infrastructure

Pages are available in English (/en) and Dutch (/nl), except the AI receptionist build guide and the 10DLC article, which are English only.

## Contact
- Email: info@voicedock.ai
- Phone: +31 570 238 200
- Leeuwenbrug 89a, 7411 TH Deventer, Netherlands (EU)