What 'beta' actually means here
This page is deliberately honest: the AI Platform is in beta, which means it is offered for evaluation and prototyping, not as a finished product with a service-level agreement. Features, model availability, and behavior can change while it matures, and access is enabled per account rather than opened to everyone at once. Treating it as an experiment you can influence — rather than a stable dependency — is the right mental model for now.
Controlled early access, on purpose
Access is granted through a review and a support-assisted activation instead of an instant self-serve switch. That is intentional: a controlled rollout lets the platform match real usage to available capacity, catch rough edges with a smaller group, and give early users direct support while things are still moving. The trade-off is that you request access and describe your use case rather than simply flipping it on, which also means the guidance you get is tailored to what you are trying to do.
Token-based API access
Integration is done with an API token issued under your cdn.com.tr account. The token authenticates your application's calls to the AI API, and you manage its lifecycle from the panel — issue it, reset it if it may have leaked, and revoke it when a project ends. As with every credential on the platform, keep it server-side in an environment variable or secret store and never ship it in a browser bundle or commit it to a repository.
No capacity or model guarantees during beta
Because this is beta, there is no fixed capacity promise and no locked model guarantee. That is not a hidden footnote — it is the deal, and it is why the use case is clarified with support rather than sold as a tier. For you it means designing for graceful degradation: if an AI call is slow or unavailable, your application should fall back sensibly rather than breaking, and you should avoid putting a beta dependency on the critical path of a revenue flow.
Where it fits and where it does not
Beta is a good fit for internal tools, proofs of concept, non-critical enrichment, and experiments where you want to learn whether an AI-assisted feature is worth building at all. It is a poor fit for anything that needs guaranteed uptime, strict latency, or contractual commitments today, because those are exactly the assurances beta does not yet make. Choosing the right first project keeps your evaluation honest and low-risk.
The path from beta to production
The point of trying something in beta is to make a confident production decision later. Keep your integration loosely coupled and observable so you can judge quality, latency, and cost against real requests, and keep notes you can take back to support. When you are ready to depend on it, that conversation is where scope, capacity expectations, and any commitments get set — so nothing about a live rollout is assumed from the beta experience alone.
How to set it up, step by step
Request early access
From the panel, open the AI Platform area and request access, or raise it with support. Because this is a controlled beta rather than instant self-serve, activation is reviewed and enabled for your account rather than granted automatically. Describe what you want to build so the right scope is turned on.
Describe your use case with support
Share the shape of your workload — roughly what kind of requests, how often, and what a success looks like. During beta this conversation sets expectations on both sides, because there is no fixed capacity tier you are simply billed into. It also helps flag early whether your idea is a good fit for the current beta surface.
Issue an API token
Once activated, generate an API token in the panel. Treat it as a secret: store it in your secret manager or as an environment variable, never in client-side code or a public repo. The token authenticates your calls to the AI API and can be reset if you suspect exposure.
Integrate and test with limited scope
Wire the token into a test or staging environment first and send representative requests. Keep the blast radius small during beta — a feature flag or a single non-critical workflow — so you can evaluate quality and latency without exposing customers to something that may change.
Watch status and iterate
Use the panel status view to see platform state and confirm your integration is reachable. Because behavior can evolve during beta, keep your integration loosely coupled and be ready to adjust as the service matures. Feed observations back to support — beta feedback shapes what ships.
Plan the production conversation early
If the evaluation goes well, talk to support about what a supported, production-ready arrangement would look like before you depend on it. This is where scope, capacity expectations, and any commitments get agreed, rather than being assumed from the beta experience.
Example scenarios
A product team wires the AI API into a staging build behind a feature flag to judge whether an AI-assisted feature is worth the engineering investment, without exposing customers.
A team uses beta access for an internal-only helper — where occasional changes or downtime are acceptable — to get real experience with the API before any customer-facing plan.
An application adds optional AI-generated summaries or tags as an enhancement, designed to degrade gracefully if a call is slow or unavailable, so the core experience never depends on it.
A developer validates the end-to-end flow — issue a token, authenticate, send requests, watch status — to confirm the integration pattern before planning a larger rollout with support.
Frequently asked questions
Is the AI Platform generally available?
No. It is in beta and offered for early access. Activation is enabled per account through a support-assisted process rather than an instant self-serve signup, and features may change while the platform matures.
Are there capacity, uptime, or model guarantees?
No fixed capacity promise, SLA, or locked model guarantee is made during beta. That is why the use case is clarified with support up front. Design your integration to degrade gracefully and keep it off the critical path of revenue-bearing flows.
How do I get access?
Request early access from the AI Platform area in the panel or raise it with support, and describe what you want to build. Access is reviewed and enabled for your account, which also lets support turn on the right scope for your case.
How is the API secured and how do I manage the token?
Access is token-based. Issue an API token in the panel, keep it server-side in an environment variable or secret store, and reset or revoke it if it may be exposed or when a project ends. Never place it in client-side code or a public repository.
Should I put a beta AI feature in production traffic?
Not on the critical path. Beta is meant for evaluation and prototyping, so favor internal tools, proofs of concept, and non-critical enrichment. When you want a supported production arrangement, agree scope and expectations with support first.
Will my beta integration keep working as it evolves?
It may change, since behavior and available models can move during beta. Keep the integration loosely coupled and observable so you can adapt, and feed observations back to support — beta feedback directly influences what the platform ships next.