Service plans and limits

Account state and provider documentation were verified 13 September 2026. Provider quotas and product names can change; recheck the linked primary source and the relevant dashboard before every release that depends on a limit.

EXEPERT currently uses free service tiers for the web platform. There is no active paid runtime, paid inference fallback, trial, or automatic upgrade in the paths documented here. Domain registration is a separate account expense and is outside these runtime quotas.

Production service posture

Provider	Active plan and charge posture	EXEPERT modules	Material limits used for planning	Behavior at the limit
Vercel	Hobby, $0. No Hobby billing cycle. Existing production use must continue to satisfy Vercel's personal/non-commercial fair-use terms; this page does not certify eligibility.	exepert serves app.exepert.com, its API, and /docs; exepert-landing serves the website/blog.	4 active CPU hours, 360 GB-hours provisioned memory, 1,000,000 function invocations, 100 deployments/day, and 50 domains/project.	Hobby resources pause or remain unavailable until their reset window. EXEPERT has no on-demand paid overage path. Static and dynamic behavior can therefore fail at different times.
Cloudflare	Workers Free / Pages Free, $0.	Brain Learning app and redirect, identity gateway, Pages Functions, Turnstile, and the disabled Workers AI tutor path.	100,000 Worker requests/day; 10 ms CPU/invocation; 20,000 static files/version; 25 MiB per static file. Workers AI provides 10,000 neurons/day at no charge.	Dynamic function or AI calls fail after quota; the public static lessons and device progress remain usable. Tutor returns authored material and never routes to a paid provider.
Supabase	Free, $0.	Shared PostgreSQL data plane, RLS, Data API, Realtime, Storage, Brain progress schema, and Edge identity candidate.	2 active projects; 500 MB database/project; 1 GB storage; 5 GB egress plus 5 GB cached egress; 50,000 MAU; 500,000 Edge Function invocations. Free projects may pause after one week of inactivity. Automatic backups and point-in-time recovery are not included.	Hosted identity/data features can become unavailable or pause. AI Lab offline behavior, public blog fallback handling, and static Brain Learning remain independent; no paid project is created automatically.
Hugging Face	Free personal account, $0.	Public Open Brain Kit dataset and static Brain Learning demonstration Space.	Public Hub storage and static Space are used. CPU Basic exists as a free compute tier, but the production demonstration is static and does not rely on hosted inference or GPU capacity.	Downloads or the demonstration may be temporarily unavailable under Hub controls. The canonical learning app retains its bundled lessons; no paid endpoint or ZeroGPU dependency is used.
Resend	Free, $0.	Better Auth verification and password-reset email from the active identity service.	3,000 emails/month, 100 emails/day, 3 sending domains; free-plan overage is unavailable.	New verification/reset mail stops until quota resets. Existing sessions, public pages, and local learning remain available. There is no automatic paid send.
Google Cloud	No billing account attached to the OAuth project; $0 runtime use.	Existing Google OAuth client for Better Auth social sign-in. No paid Google API is part of the documented runtime.	OAuth consent/client configuration is the only active dependency recorded here. Any future billable Google API requires separate review and billing authorization.	Google sign-in fails closed if its client, callback, or provider is unavailable. Email/password and public unauthenticated experiences remain separate.
GitHub	GitHub Pro entitlement with a current $0 account charge through an account benefit. No expiry or benefit details are recorded here.	Private 4kmal/MASTER-exe repository, pull requests, reviews, Actions CI, and gated Cloudflare/Supabase workflows.	GitHub Pro includes 3,000 Actions minutes/month and 1 GB Actions artifact/cache storage for private repositories.	CI and manual deployment workflows wait or fail when included usage is unavailable. Production continues at its last deployed commit; no paid runner capacity is enabled automatically.

Provider notes and primary sources

Vercel

The Vercel Hobby plan is free and applies monthly included usage, with a non-commercial personal-use fair-use boundary. The general limits reference contains framework and platform ceilings that can differ from included usage. EXEPERT uses two projects so the AI Lab/API/docs and website/blog can deploy and roll back independently.

Cloudflare

Workers pricing defines the free request and CPU allocation, while Workers limits defines static asset and runtime ceilings. The Workers AI pricing page states that the free allocation resets daily and excess use on Workers Free fails rather than becoming paid inference.

The tutor remains disabled, so the current production site consumes no Workers AI request during normal learning. Turnstile and the identity gateway have separate configuration gates; a working static Pages response is not evidence that either is enabled.

Supabase

The Supabase pricing page is the authority for the Free project limits. The data plane holds small relational records and publish-safe media metadata. MRI archives and the downloadable Brain Kit do not belong in Supabase Storage. Free-tier absence of automatic backups is recorded as an operational constraint; this documentation release performs no backup or database action.

Hugging Face

Hugging Face billing explains free and usage-based Hub services. The Spaces overview lists the available hardware tiers. EXEPERT uses a static Space and a public dataset, with no Inference Endpoint, paid accelerator, or always-on ML runtime.

Resend

The Resend pricing reference defines the daily/monthly Free limits and the absence of Free overage. Resend is an identity mail transport, not an application database or a general campaign system.

Google Cloud

The OAuth project currently has no attached billing account. Google OAuth configuration must preserve both the active app.exepert.com callback and any separately gated candidate callback. See the Google Cloud billing overview before enabling a billable Google service.

GitHub

GitHub included usage defines Actions minutes and storage by plan. The private repository is 4kmal/MASTER-exe (access required). Actions deploy only reviewed commits and remain a release mechanism, not a production runtime.

Cost-control rules

• Keep identity, progress, and tutor flags independent. Enabling a static site never implies authorization for a dynamic provider.
• Do not add a paid provider, trial, card-backed overage, prepaid AI credit, or automatic plan upgrade as a reliability fallback.
• Serve large public release archives from the pinned Hugging Face dataset. Keep Vercel, Cloudflare static assets, and Supabase focused on their intended application and record sizes.
• When a quota is exhausted, preserve a truthful degraded state: public pages, device-local learning, authored explanations, and the last verified deployment should remain available where their provider is unaffected.
• Recheck dashboards without copying account emails, account IDs, billing IDs, balances, usage totals, credentials, or benefit expiry into this page.