Managing multiple AI providers can mean juggling keys, accounts, and bills. Hostinger AI Router promises to bring them together, but how well does that work? I tested two model APIs, tracked the costs, and questioned support. Read on for the results and limitations.
Tip: Start with a small credit balance, test the models your application needs, set a key expiration date, and monitor the first few requests before moving to production. Hostinger documents per-key limits, but confirm those controls in your account.
HostAdvice’s Hosting Rating Methodology explains the broader evaluation principles behind our reviews. For an AI API gateway, I adapted the categories to the decisions developers actually face: model access, integration, request performance, billing transparency, usage controls, and the quality of technical support. Scores reflect the hands-on evidence collected.
| Parameter | Score /10 | Why This Score |
|---|---|---|
| Pricing & Billing | 8.6 | Straightforward prepaid billing and accurate charges, although purchased AI credits are nonrefundable. |
| Model Access & Features | 8.7 | Extensive model selection, detailed pricing, and unified access. |
| API Performance | 8.0 | Both models responded successfully, although limited testing cannot establish consistent speed or reliability. |
| Ease of Use | 9.0 | Smooth onboarding, simple API key creation, clear navigation, and effortless switching between providers. |
| Usage Tracking & Controls | 8.4 | Accurate usage reports and useful monitoring tools, but advanced access controls remain untested. |
| Support & Documentation | 8.2 | Immediate AI assistance and helpful guides, but human review left important technical questions unresolved. |
| Overall | 8.4 | A practical multi-model gateway with transparent billing, held back by unverified reliability features and support limitations. |
Hostinger presented three prepaid credit packages in my account: 500 credits for $5, 2,000 credits for $20, and 5,000 credits for $50, all before taxes. Each package therefore had the same purchase rate of one US cent per credit; I did not see a bulk discount. Credits are not the same as a fixed number of requests.
Consumption depends on the selected model and the input and output tokens processed, so the model catalogue and activity logs matter more than package size alone. The checkout added 16% VAT to my $5 selection, bringing the total to $5.80. That tax rate reflects this order, not necessarily every customer’s bill.
Credit packages shown during my test:
| Credits | Price (excl. tax) | Unit price |
|---|---|---|
| 500 | $5.00 | $0.01 |
| 2,000 | $20.00 | $0.01 |
| 5,000 | $50.00 | $0.01 |
Hostinger’s pricing page advertises usage-based billing rather than a fixed monthly charge. That is appealing for irregular workloads, but the real value depends on the token rates for the models you choose.
I later compared two live API charges with the model rates and found both calculations accurate. That establishes transparent billing for these small requests, not the relative value of every available model.
Hostinger supports several payment options. During checkout, I found:
The checkout also included a coupon-code field and displayed taxes before payment. When I selected the 500-credit package, Hostinger added 16% VAT, bringing my total to $5.80. Your final amount may differ depending on applicable taxes.
There are two important conditions to consider before purchasing credits.
First, I did not encounter an AI Router-specific free trial, and Hostinger’s setup documentation does not advertise a standard trial.
Second, AI credits are nonrefundable. Although Hostinger promotes a 30-day money-back guarantee for eligible hosting services, its refund policy explicitly excludes LLM/AI credits. You should therefore avoid purchasing more than you initially need.
For the initial onboarding test, I used Hostinger’s public Get started button, signed in to hPanel, reviewed the credit packages and checkout, and opened the AI Router management area.
I recorded the account steps, advertised functionality, purchase amounts, taxes, payment choices, and dashboard navigation. I did not treat promotional feature descriptions as proof that the features work.
After checkout, I sent one authenticated request to DeepSeek V4.1 Flash and one to Claude Haiku 4.5 using the same endpoint and key. I captured HTTP status, elapsed time, output completion, token counts, credit deductions, and the Activity page’s totals. This gives a limited but reproducible test of routing and accounting, not a long-term uptime benchmark.
A router is only useful if it can accept an application’s request, send it to the selected model, and report what happened. I tested Hostinger directly from my terminal using curl and its OpenAI-compatible chat-completions endpoint.
I sent the same short question to two providers through the same URL and API key, changing the model identifier and output-token limit.
I recorded each HTTP status, total elapsed time, response status, and usage record.
I started with DeepSeek V4.1 Flash. Hostinger returned HTTP 200 in 3.170384 seconds, identified the requested model correctly, and reported 41 input tokens and 100 output tokens.
The response was cut short with finish_reason set to length. Its usage details showed 74 reasoning tokens within the output allowance, so the truncated explanation was caused by my 100-token cap, not an API-routing error.
Alt text: Terminal showing a curl request to Hostinger AI Router using DeepSeek V4.1 Flash, HTTP 200, 3.170384-second response time, and 0.0090024 credits charged.
Next, I sent the same question to Claude Haiku 4.5 using the same credentials and endpoint. This time I allowed up to 200 output tokens.
Claude returned HTTP 200 in 1.523767 seconds and finished its answer normally, using 18 input tokens and 37 output tokens. That confirms the practical model-switching workflow: I did not create a second provider account or change the gateway URL.
Alt text: Terminal showing a successful Claude Haiku 4.5 API request through Hostinger AI Router, HTTP 200, 1.523767-second response time, and 0.02436 credits charged.
The DeepSeek request took 3.17 seconds end to end, compared with 1.52 seconds for Claude.
These are individual observed request times, not averages or router-only latency measurements. The models used different token budgets and generated different amounts of output.
With only one successful request per model, I cannot claim that Claude is consistently faster or estimate uptime. Both API routes worked on the first authenticated attempt.
I compared the exact charges reported in each JSON response with the rates displayed in Hostinger’s model catalogue. DeepSeek reported 0.0090024 credits, calculated from 41 input tokens at 26.4 credits per million and 100 output tokens at 79.2 credits per million.
Claude reported 0.02436 credits, calculated from 18 input tokens at 120 credits per million and 37 output tokens at 600 credits per million. Both amounts matched those calculations.
The Activity page independently showed two prompts, 196 tokens, no cached tokens, and model-level entries of 141 tokens for DeepSeek and 55 for Claude. The API charged 0.0333624 credits in total; hPanel rounded this to 0.03 credits.
The Overview balance likewise displayed 2,047.97 credits after the test. Its summary matched the rounded activity numbers, so I found no discrepancy in the usage records for these two requests.
Alt text: Hostinger AI Router Activity page showing two prompts, 196 total tokens, 0.03 displayed credits used, zero cache, and model-level token counts for DeepSeek and Claude.
Alt text: Hostinger AI Router Overview dashboard showing 2,047.97 available credits, two prompts, 196 tokens, 0.03 rounded credits used, and zero cache.
Hostinger AI Router performed well in my tests. Both DeepSeek V4.1 Flash and Claude Haiku 4.5 returned successful responses through the same API key and endpoint, taking 3.17 and 1.52 seconds, respectively. The reported token usage and credit deductions also matched the published rates and reconciled with the dashboard totals.
For you, these results mean:
The tests confirmed that Hostinger’s model switching and usage tracking work as advertised for these requests. However, the small sample doesn’t establish long-term reliability, and automatic fallback and caching savings still require verification.
I began on the AI Router product page, where the main call to action was easy to find.
Alt text: Hostinger AI Router landing page showing the Get started button, single API key access, supported AI models, and key platform features.
Clicking Get started sent me to Hostinger’s account sign-in page rather than directly to an API setup form. The sign-in screen offered Google, Facebook, and GitHub buttons alongside email and password fields. Existing Hostinger customers can therefore start with the account they already use.
Alt text: Hostinger sign-in form showing Google, Facebook, GitHub, email, password, and Log in controls.
After signing in, I was taken to Hostinger’s AI Router credit top-up page in hPanel. There was no separate installation or setup wizard.
Instead, the screen presented the three credit packages alongside the product’s main selling points, with a Purchase credits button below. Since I wanted to test the service without spending much upfront, I proceeded with the 500-credit package.
Alt text: Hostinger AI Router credit top-up page showing 500 credits for $5, 2,000 for $20, and 5,000 for $50, alongside key features and the Purchase credits button.
Clicking Purchase credits opened an Add credits dialog where I confirmed my 500-credit selection. Hostinger showed a $5.00 subtotal, added $0.80 in taxes and fees, and displayed a $5.80 total.
There was also an option to enter a coupon code. From there, I clicked Choose payment method to continue to the checkout page, so I could see the complete charge before entering payment details.
Alt text: Add credits modal showing Hostinger Credits 500, a $5.00 subtotal, $0.80 taxes and fees, and a $5.80 total.
The payment page visibly offered bank card, PayPal, Google Pay, and Apple Pay, plus an expandable Additional payment methods section.
The card option displayed Visa, Mastercard, American Express, and Discover branding. I could inspect the order summary without selecting a payment method first.
Alt text: Hostinger checkout page showing card, PayPal, Google Pay, Apple Pay, additional payment methods, and a $5.80 order total.
I expanded the Card option to examine the form. It requested the cardholder name, card number, expiration date, and CVC, followed by a Submit payment button. After payment, I received an email confirmation and Hostinger took me to AI Router in hPanel. The card form also mentioned recurring payments, so I would check the account’s payment settings rather than assume future top-ups are automatic or rule them out.
Alt text: Hostinger card checkout showing name on card, card number, expiry, CVC, Submit payment, and the 500-credit order summary.
Verdict: Pricing and Checkout
The upfront credit packages are easy to compare, and I appreciated seeing the tax-inclusive total before payment. Having several payment methods also helps. More importantly, the two real API deductions matched the advertised token rates. Costs still vary substantially by model, so a small initial top-up does not by itself guarantee low spending at scale.
With checkout complete, I arrived on the AI Router Overview page inside hPanel. That was the handoff I wanted to see: I was no longer looking at the sales or payment screens, but at the dashboard where I could manage credits and connect an application. Its layout prioritized the information I needed next: remaining credits, integration guidance, and usage activity.
The main navigation was divided into Overview, Integration, Activity, Models, and API key. That separation is helpful because each answers a different practical question: How much credit remains? How do I connect an application? What have I spent? Which models are available? How do I create or manage credentials?
The Overview page showed the available credit balance in a prominent card with Top up and Add API key controls. A second card, Start using AI Router anywhere, pointed to connection guides for OpenClaw, Hermes Agent, Cursor, and the OpenAI SDK. Below those, a seven-day summary tracked total credits used, prompts, tokens, and cache activity. Before making API requests, the usage counters shown in my session were zero.
Alt text: Hostinger AI Router Overview screen displaying available credits, Add API key, Top up, integration links, and summary metrics for credits, prompts, tokens, and cache.
Before writing any code, I opened Models to see what the gateway actually offered. The catalogue showed 149 listings, including separate regional variants for some models, rather than an undifferentiated list of 149 providers.
Search, filters, sortable columns, copyable model IDs, context windows, maximum outputs, and distinct input/output prices helped me narrow the options quickly.
Alt text: Hostinger AI Router Models table displaying Claude model variants, search and filtering, context windows, maximum output tokens, and input and output prices per million tokens.
I searched for DeepSeek and found DeepSeek V4.1 Flash, with a 1,310,720-token context window, a 393,216-token maximum output, and prices of 26.4 input credits and 79.2 output credits per million tokens.
I chose it alongside Claude Haiku 4.5, listed at 120 input and 600 output credits per million tokens, because the two models let me test different providers and different price points.
Alt text: Hostinger AI Router model search results for DeepSeek V4.1 Flash, including model ID, context window, maximum output, and input and output token prices.
Next, I opened API key and selected Add API key. The creation dialog asked for a descriptive key name and an expiration period. I named mine hostadvice-review-test and used a one-month expiration so the testing credential would not remain valid indefinitely.
Other choices were three months, six months, one year, or no expiration. Hostinger displayed a copyable secret after creation, and the key list marked the credential active with access to all models.
Alt text: Hostinger AI Router Add API key dialog containing the Key name input, Expiration dropdown, and Save button.
Alt text: Hostinger AI Router API key expiration dropdown showing one month, three months, six months, one year, and Never expires choices.
I did not see model-specific restrictions, spend ceilings, or request limits in that initial creation dialog. That alone does not prove such controls are unavailable elsewhere: the key list also has a details view, which needs checking before I reach a conclusion about access controls.
The Integration page automatically selected my new key and divided setup guidance into Apps & automation, Coding agent & IDE, and SDKs & API. Visible choices included OpenClaw, Hermes Agent, n8n, AnythingLLM, FeedCraft, Morphic, Open Notebook, Open WebUI, and SQLChat.
The choice of tools is helpful, but it would be unnecessary to install an application merely to prove that the gateway works. I therefore used a direct curl request for the model-switching test described in the Performance section.
Alt text: Hostinger AI Router Integration page with a selected API key, tabs for Apps and automation, Coding agent and IDE, SDKs and API, and cards for OpenClaw, Hermes Agent, n8n, and other tools.
The journey from Get started to live API usage was straightforward. I moved from sign-in and credit selection to the dashboard, browsed model rates, created a temporary key, and connected through curl without installing a separate application.
The menus were easy to follow, though the initial key-creation dialog did not expose spending or model restrictions; I still need to inspect the key-details controls before judging those options.
For an API gateway, support should do more than explain how to create a key. I also want clear answers about failed requests, fallback, and unexpected charges, because those issues can affect a live application.
I tested Hostinger’s chat assistance and searched its Knowledge Base for supporting technical information.
Hostinger advertises 24/7 assistance. Its support and self-service channels include:
Hostinger does not offer telephone support. I focused on in-dashboard chat and the Knowledge Base because these are the most relevant channels for technical AI Router questions.
I opened Hostinger Agent in hPanel and asked a question I had not been able to verify in my API tests: if a model becomes unavailable, does AI Router automatically choose another, and how can I identify the model used and check its charges? I sent the question at 21:19 and received the AI answer in the same minute. There was no visible queue for that initial exchange.
The reply gave me a reasonable starting point. Hostinger Agent said fallback was automatic, advised checking the model field in the API response, and suggested comparing that with usage grouped by model in hPanel. It also explained how to check charges against the shared credit balance.
However, it did not specify which errors activate fallback or exactly how a replacement model is billed. I treated those details as support explanations, not verified results.
Alt text: Hostinger Agent chat showing the original fallback and billing question and the AI response describing automatic routing and usage checks.
At 21:21, I asked to speak with a human specialist. Rather than transferring the conversation, Hostinger Agent outlined its own plan to investigate the issue and suggested continuing with the assistant.
After I requested a human again, it asked for a redacted API response and finally displayed two choices: Wait for a human review or Continue with Hostinger Agent.
Alt text: Hostinger Agent conversation showing a request for human help followed by a plan to investigate fallback and reconcile credit usage.
Alt text: Hostinger Agent showing Wait for a human review and Continue with Hostinger Agent buttons after another escalation request.
I selected human review and supplied four specific questions at 21:22. I wanted to know if timeouts and rate limits trigger fallback, if customers can select replacement models, which model’s rates apply, and if individual fallback requests can be identified. The interface preserved the conversation and showed a review status with Jenevie’s name, so I did not need to start from scratch.
Alt text: Hostinger chat displaying the four technical fallback questions and a status card indicating that Jenevie was reviewing the request.
A response appeared at 21:29, approximately eight minutes after escalation. It was marked Response reviewed by Jenevie. I appreciate the transparency, but this was visibly a human-reviewed reply rather than a conventional back-and-forth conversation directly with a specialist.
Unfortunately, the response did not answer my four questions. Instead, it linked to the AI Router getting-started guide, its help category, an OpenClaw credits article, and a guide for changing AI providers in Docker.
The links were related to AI Router, but the answer did not establish the fallback triggers, the charge applied after rerouting, user control over replacement models, or request-level fallback records. It concluded by inviting me to ask again if those articles did not help.
Alt text: Hostinger response marked reviewed by Jenevie at 21:29, containing four links to AI Router and Docker documentation.
Verdict: Live Chat
Hostinger Agent was immediate, easy to approach, and useful for the basics. I could eventually request human review, and the original context survived the handoff. However, reaching that stage required repeated requests, and the roughly eight-minute wait ended without a technical resolution.
This was acceptable for initial guidance but disappointing for questions that affect fallback behaviour and production costs.
I then opened Hostinger’s Help Center. Its homepage uses a prominent search field and clearly labelled categories for products such as API, Billing, VPS, Hostinger Agent, Domains, and other services.
The layout felt clean, and the category cards made it easy to start browsing without knowing a particular article title.
Alt text: Hostinger Help Center homepage showing the main search field, suggested topics, and cards for dashboard, AI Builder, API, domains, DNS, and other services.
Searching for AI router returned 28 results in my session. Titles and brief descriptions were easy to scan, but the results also included broader Docker, domain, and VPS articles. That is useful for related configuration work, though less efficient when you need one precise answer about AI Router’s gateway behaviour.
Alt text: Hostinger Knowledge Base search results for AI router, showing 28 matching articles with titles and short descriptions.
The dedicated AI Router help category was much narrower, with three articles: a getting-started guide, a nexos.ai credit migration guide, and an OpenClaw credits guide. The getting-started article was the most useful general reference. It covered the API URL, authentication, example requests, model switching, shared credits, and documented key-level spending and model restrictions.
I also examined the nexos.ai migration guide. It uses a before-and-after explanation, walks readers through credit conversion and API configuration updates, and includes troubleshooting steps. That made it more practical than a generic product overview, especially for customers moving an existing application.
Alt text: Hostinger Help Center guide about migrating nexos.ai credits, including a before-and-after comparison and instructions for updating API settings.
The gap was advanced fallback documentation. I could find the general claim that AI Router reroutes requests when a model is unavailable, but not a precise explanation of timeout and rate-limit behaviour, replacement-model billing, customer-selected fallback models, or request-level fallback auditing. The human-reviewed response did not supply that missing information.
Verdict: Knowledge Base
The Help Center is easy to search and its stronger AI Router guides provide practical setup, migration, and billing instructions. For routine integration questions, I found it useful. However, the dedicated AI Router collection remains limited, and the material I checked did not fully address failure scenarios that developers may need to understand before deploying critical workloads.
Hostinger’s support impressed me with its quick initial response and accessible documentation, but the experience became less convincing when I needed detailed technical guidance. Kodee answered immediately, while the human-reviewed response took approximately eight minutes and still left my main questions about automatic fallback and billing unresolved.
Here’s what my testing means for you:
Overall, I found Hostinger’s support better suited to onboarding and routine troubleshooting than resolving complicated API issues.
The quick AI assistance and organized Knowledge Base are definite strengths, but the lack of a clear resolution after human review was disappointing. If you’re building a production application, I’d recommend getting explicit clarification on fallback behavior and billing before relying on those features.
Yes. I recommend Hostinger AI Router for developers, businesses, and AI enthusiasts who want access to multiple AI models without managing separate provider accounts, API keys, and billing systems.
My hands-on tests confirmed its main advantage: I successfully connected to DeepSeek and Claude through the same API endpoint, with accurate credit deductions and useful usage reports. The straightforward setup, extensive model catalogue, and flexible prepaid billing make it particularly appealing for experimenting with AI models and building applications.
However, I wouldn’t recommend relying on its automatic fallback for mission-critical applications without further verification. Although Hostinger advertises this capability, I couldn’t test it directly, and human-reviewed support failed to clarify important details about fallback conditions and billing.
| Description | Expert Review |
|---|---|
| Budgetvriendelijke hosting met hoge prestaties en eenvoudige beheertools. | Read Shared Hosting Review |
| ast en veilige WordPress-hosting met installatie met één klik en premiumfuncties. | Read Wordpress Hosting Review |
| Schaalbare VPS-hosting met toegewijde resources en roottoegang. | Read VPS Review |
| Snelle, flexibele cloudhosting met uitstekende beschikbaarheid en schaalbare hulpbron... | Read Cloud Hosting Review |
| Veilige en privé-hostingoplossingen met offshore datacenterlocaties. | Read Offshore Hosting Review |
| Veilige en betrouwbare e-mailhosting met professionele functies. | Read Email Hosting Review |
| Betrouwbare Python-hosting met flexibele omgevingen voor ontwikkelaars. | Read Python Hosting Review |
| High-performance PHP-hosting met volledige ondersteuning voor dynamische websites en ... | Read PHP Hosting Review |
| Betrouwbare Windows VPS-hosting met volledige controle en aanpassingsmogelijkheden. | Read Windows VPS Review |
| Snelle en flexibele hosting op maat voor Node.js-toepassingen met optimale prestaties... | Read Nodejs Hosting Review |
| Geoptimaliseerde hosting voor WooCommerce-winkels met hoge snelheid en beveiligde int... | Read Woocommerce Hosting Review |
| Dedicated serverhosting voor naadloze Minecraft-gamingervaringen. | Read Minecraft Server Hosting Review |
| Schaalbare hostingoplossingen met geavanceerde functies voor digitale bureaus en ontw... | Read Agency Hosting Review |
| Snelle, beveiligde hosting geoptimaliseerd voor Magento e-commercewebsites. | Read Magento Hosting Review |
| Hoogpresterende Linux-gebaseerde hosting voor stabiele en veilige website-operaties. | Read Linux Hosting Review |
| Robuuste Java-hostingoplossingen voor dynamische webapplicaties en projecten. | Read Java Hosting Review |
| Geoptimaliseerde hosting voor e-commercewebsites met veilige, snelle en betrouwbare p... | Read Ecommerce Hosting Review |
| Betrouwbare Django-hosting met hoge snelheden en een veilige omgeving. | Read Django Hosting Review |
| Gebruiksvriendelijke cPanel-hosting met robuuste prestaties en betrouwbare ondersteun... | Read Cpanel Hosting Review |
| Krachtige hosting voor bedrijven met hoge snelheden, beveiliging en schaalbaarheid. | Read Business Hosting Review |
| Easy-to-use website builder with drag-and-drop tools and customizable templates. | Read Website Builder Review |
| Optimized hosting for Joomla sites with one-click installation and reliable performan... | Read Joomla Hosting Review |
| Powerful hosting with full PostgreSQL database support for data-driven applications. | Read PostgreSQL Hosting Review |
| Flexible hosting with MongoDB integration for scalable, modern web applications. | Read MongoDB Hosting Review |
| AI-powered website creation platform for building professional sites in minutes. | Read AI Builder Review |
| Reliable hosting for n8n workflow automation with easy setup and management. | Read n8n Hosting Review |
| VPS hosting with Docker support for containerized application deployment and scaling. | Read Docker VPS Review |
| Toegewijde SMTP-serverhosting voor betrouwbare en veilige e-mailbezorging. | Read SMTP Server Review |
| Snel en geoptimaliseerd hosting op maat van Ruby on Rails-webapplicaties. | Read Ruby on Rails Review |
| Functierijke hosting met OpenClaw-integratie voor het bouwen en beheren van claw mach... | Read OpenClaw Review |
| Snelle en betrouwbare hosting met in het VK gevestigde servers voor optimale lokale p... | Read UK Hosting Review |
| Betaalbare en betrouwbare hosting met servers in India voor toegang met lage latentie... | Read India Review |
| Read Singapore Review | |
| Read Australia Review | |
| Read AI Agent Review | |
| Read Paperclip VPS Review | |
| Read Hermes Agent Review | |
| Read Hostinger Connector Review | |
| Read Web Apps Hosting Review | |
| Read Hostinger Reach Review | |
| Read MCP Review | |
| Read hpanel Review | |
| Read Odoo Review | |
| Read Laravel Review | |
| Read MERN VPS Review | |
| Read Ubuntu Review | |
| Read Drupal Hosting Review | |
| Read Express.js Review | |
| Read React Review | |
| Read Nextjs Review | |
| Read GPU Hosting Review |
Hostinger AI Router is an API gateway that connects applications to supported AI models from multiple providers through one API key and a shared credit balance.
No separate hosting plan is described as a requirement for direct API use. You need a Hostinger account, an API key, and available credits; the gateway can serve applications hosted elsewhere.
I did not find a dedicated AI Router free trial or guaranteed free credit allowance. Direct API requests require available credits, although an existing Hostinger account may already have a balance.
It deducts prepaid credits according to the selected model’s input and output token rates. There is no mandatory monthly subscription, and usage appears in hPanel.
Hostinger advertises automatic rerouting if a model is unavailable. I did not observe a fallback event, and support did not clarify all the triggering conditions or replacement-model billing rules.
Beantwoord een paar eenvoudige vragen en vind de perfecte oplossing voor jou!
Hosting zoeken starten
HostAdvice.com biedt professionele web hosting beoordelingen aan, compleet onafhankelijk van andere bedrijven. Onze beoordelingen zijn onpartijdig, eerlijk en de evalutieprocedure is hetzelfde voor alle beoordelingen.
Een compensatie wordt verkregen van de bedrijven die we beoordelen. Compensatie van diensten en producten hebben geen invloed op de richting of conclusies van onze beoordelingen. De compensatie beïnvloed onze ranglijsten voor bepaalde hostbedrijven ook niet.
Deze compensatie bestaat uit de kosten van royalty's voor beoordelaars, voor het aanschaffen van de accounts en voor het testen.
