COST ARCHITECTURE
See what can launch without immediate software overhead.
A launch allowance is a provider’s documented allocation or credit in a client-owned account. It can reduce early operating cost. It does not reduce the strategy, design, engineering, security, testing, or verification required to build the system.
FILTER THE OPERATING MAP
Start with the capability. Open the technical record only when useful.
The summary shows the current boundary and review state. Expand any record for consumption, privacy, fallback, and upgrade details. This keeps a complex operating map readable without hiding the evidence.
Serverless application logic100,000 Worker requests per day; static asset requests should bypass Worker execution.
WHY THIS MATTERS
Workers Free request allocation for dynamic functions. This is what runs the site's dynamic behavior—forms, routing, live logic—without renting a server that costs money whether or not anyone visits.
Illustrative low-volume operation within the current documented allowance: A local service business gets around 3,000 visits a month and roughly 40 contact-form submissions. That is a few thousand Worker requests a day at most—comfortably inside the 100,000/day allowance, so the site runs with no hosting-compute bill.
Expected light use: A static-first marketing site with low-volume forms and optional module requests.
- What consumes it
- Dynamic Worker or Pages Functions invocations. Static asset requests do not consume this request allocation.
- Fallback
- Static content and conventional navigation remain available.
- Upgrade trigger
- Dynamic requests approach 70% of the daily allocation or CPU limits constrain execution.
- Likely paid successor
- Cloudflare Workers Paid
- Commercial use
- Commercial use permitted under provider terms.
- Data boundary
- Review logs and binding data; do not send unnecessary sensitive data.
Spam and abuse protectionProvider-managed free plan; server-side token validation required.
WHY THIS MATTERS
Production-suitable free Turnstile plan for most small and medium sites. Keeps bots and spam out of contact and inquiry forms without forcing real customers to solve an annoying puzzle before they can reach the business.
Illustrative low-volume operation within the current documented allowance: A consulting firm gets 25–50 genuine form submissions a month plus a similar volume of bot attempts. Turnstile screens all of it on Cloudflare's free, unlimited-volume plan—there is no submission count that pushes a small business off this tier.
Expected light use: Abuse protection for contact and project-brief forms at typical small-business traffic.
- What consumes it
- Widget challenges and token verification requests under the provider-managed free plan.
- Fallback
- Honeypot, rate limit, input constraints, and manual contact path.
- Upgrade trigger
- Enterprise policy, support, or specialized controls are required.
- Likely paid successor
- Cloudflare enterprise agreement where required
- Commercial use
- Commercial use permitted.
- Data boundary
- Document privacy disclosure and validate tokens server-side.
Transactional email3,000 emails/month, 100/day, one custom domain, one webhook endpoint, 30-day retention.
WHY THIS MATTERS
Launch allowance for low-volume transactional email. Delivers the emails a business actually needs—inquiry confirmations, owner alerts—without paying for a full marketing-email platform to send a handful of transactional messages.
Illustrative low-volume operation within the current documented allowance: A home-service business gets 60 qualified leads a month and sends one confirmation to the customer plus one alert to the owner for each: 120 emails a month, well under the 3,000/month and 100/day caps.
Expected light use: Transactional visitor confirmations and owner notifications for a low volume of qualified inquiries.
- What consumes it
- Every sent or received email counts toward the daily and monthly quota.
- Fallback
- Store the structured record and show a manual follow-up path; retry through queue.
- Upgrade trigger
- Daily or monthly volume approaches 75%, longer retention is needed, or multiple domains are required.
- Likely paid successor
- Resend paid plan or alternate email adapter
- Commercial use
- Commercial use permitted subject to current terms.
- Data boundary
- Avoid sensitive content in email; redact logs.
Structured data5M rows read/day, 100K rows written/day, 5 GB account storage, up to 10 free databases, 500 MB/database, seven-day point-in-time recovery.
WHY THIS MATTERS
Low-volume SQL storage for leads, metadata, and module state. Stores leads, form submissions, and content data reliably, the same way a traditional database would, without a recurring database-hosting bill.
Illustrative low-volume operation within the current documented allowance: A professional-services firm logs 200 leads and a few hundred content records a month. That is a tiny fraction of the 5M reads/day and 100K writes/day allowance, and nowhere near the 5 GB storage ceiling.
Expected light use: Structured leads, consent records, content metadata, and low-volume module status.
- What consumes it
- Rows read or scanned, rows written, and stored database data; unindexed scans consume more rows.
- Fallback
- Queue writes, retain a minimal encrypted export, or degrade to email-only intake.
- Upgrade trigger
- Row scans, storage, recovery, or support requirements exceed the allocation.
- Likely paid successor
- Cloudflare D1 paid usage or approved SQL provider
- Commercial use
- Commercial use permitted.
- Data boundary
- No payment cards, passwords, government identifiers, or sensitive records by default.
Object storage10 GB-month, 1M Class A operations, 10M Class B operations monthly; no R2 internet-egress charge; excludes Infrequent Access.
WHY THIS MATTERS
Storage for approved media, resources, screenshots, and deliverables. Hosts downloadable resources, images, and files the business wants to share, without the surprise bandwidth fees most cloud storage charges when people download them.
Illustrative low-volume operation within the current documented allowance: A firm hosts 200 MB of PDFs, case-study images, and brand assets, with 5,000 downloads a month. That is far under the 10 GB storage cap and the monthly operation limits, and R2 does not charge for the bandwidth those downloads use.
Expected light use: Approved downloads, monitoring screenshots, and a limited library of client-owned media.
- What consumes it
- Standard-storage GB-months plus Class A mutation and Class B read operations.
- Fallback
- Disable optional uploads and preserve local/documented delivery path.
- Upgrade trigger
- Storage or operation thresholds approach 75%, or security requirements need another service.
- Likely paid successor
- Cloudflare R2 paid usage or approved storage adapter
- Commercial use
- Commercial use permitted.
- Data boundary
- Private buckets by default; signed access and retention controls required.
Read-heavy state and cache100K reads/day, 1K writes/day, 1K deletes/day, 1K lists/day, 1 GB storage.
WHY THIS MATTERS
Fast read-heavy configuration and cached data. Makes frequently-viewed content and settings load instantly by caching them at the edge, close to the visitor, instead of recalculating them on every request.
Illustrative low-volume operation within the current documented allowance: A site caches its navigation config, feature flags, and a redirect map, each read on every page view. Even at 5,000 daily visitors that stays well under 100,000 reads/day, with only a handful of writes when the owner updates a setting.
Expected light use: Feature flags, public configuration, redirect maps, and reusable cached results.
- What consumes it
- Key reads, writes, deletes, list operations, and stored data.
- Fallback
- Read from typed repository or D1 source of truth.
- Upgrade trigger
- Write volume or consistency requirements exceed KV's design.
- Likely paid successor
- Workers KV paid usage or D1
- Commercial use
- Commercial use permitted.
- Data boundary
- Eventually consistent; not for critical transactional records.
Asynchronous workflow10,000 operations/day and 24-hour message retention; a typical successful delivery uses three operations.
WHY THIS MATTERS
Move slow or failure-prone tasks out of the request path. Lets slow tasks—like sending a notification or generating a report—happen in the background so a visitor's page never hangs waiting for them.
Illustrative low-volume operation within the current documented allowance: Each of roughly 50 monthly form submissions triggers a queued email-and-logging job, using about three operations each—150 operations a month against a 10,000/day allowance, leaving enormous headroom.
Expected light use: Low-volume email delivery, indexing, report creation, and retryable background work.
- What consumes it
- Each 64 KB write, read, and delete operation; a typical successful message consumes three operations.
- Fallback
- Process synchronously only when safe, or store a retryable pending record.
- Upgrade trigger
- Operations approach 75%, retention is insufficient, or higher guarantees are required.
- Likely paid successor
- Cloudflare Queues paid usage
- Commercial use
- Commercial use permitted.
- Data boundary
- Use idempotency, retry limits, and dead-letter behavior.
AI governance and routingCore features free; up to 10 gateways and 100,000 stored logs across the account.
WHY THIS MATTERS
Analytics, caching, rate limiting, retries, and fallback for AI calls. Adds a safety and cost-control layer in front of any optional AI feature, so usage is logged, capped, and can be switched off—no runaway AI bill.
Illustrative low-volume operation within the current documented allowance: A business runs one AI-assisted module logging a few hundred requests a month. That is far under the 100,000 stored-log ceiling, and gateway routing and caching are free regardless of volume at that scale.
Expected light use: Governance, logging, caching, and limits for low-volume optional AI features.
- What consumes it
- Stored logs across gateways; model inference and guardrail usage are accounted for separately.
- Fallback
- Deterministic workflow, cached response, conventional search, or contact path.
- Upgrade trigger
- Log volume, policy, support, or enterprise governance exceeds the free plan.
- Likely paid successor
- Cloudflare paid/enterprise services or provider-neutral gateway
- Commercial use
- Commercial use permitted.
- Data boundary
- Redact prompts and avoid sensitive data in unpaid providers.
Low-volume inference10,000 neurons/day; model availability varies and some models require Workers Paid.
WHY THIS MATTERS
Optional inference for classification, summarization, and query rewriting. Powers optional AI features—like summarizing a form submission or classifying an inquiry—without a separate AI vendor contract.
Illustrative low-volume operation within the current documented allowance: A module classifies and summarizes 30–50 inbound inquiries a day. Typical classification and summarization tasks at that volume use well under the 10,000 daily neuron allowance.
Expected light use: Low-volume classification, summarization, metadata, and query rewriting.
- What consumes it
- Model-specific neuron consumption, resetting daily at 00:00 UTC.
- Fallback
- Deterministic output or retrieval-only result.
- Upgrade trigger
- Neuron use reaches 75%, selected model moves to paid, or quality requirements exceed available models.
- Likely paid successor
- Workers Paid or approved generation adapter
- Commercial use
- Commercial use permitted for approved models under current terms.
- Data boundary
- Model registry and policy checks are mandatory.
Semantic retrieval30M queried vector dimensions/month and 5M stored vector dimensions.
WHY THIS MATTERS
Prototype and low-volume vector storage and query capacity. Enables search that understands meaning, not just keywords, for site content or help articles, without a dedicated search vendor.
Illustrative low-volume operation within the current documented allowance: A knowledge base of 200 articles is indexed once, well under the 5M stored-dimension cap, and queried a few hundred times a month by an AI search module—far under the 30M monthly queried-dimension allowance.
Expected light use: A small public-content index supporting prototype or low-volume semantic retrieval.
- What consumes it
- Stored vector dimensions and queried vector dimensions across the account.
- Fallback
- Keyword search, D1 metadata filters, and curated navigation.
- Upgrade trigger
- Dimension use, corpus size, latency, or quality requires a paid vector store.
- Likely paid successor
- Vectorize paid usage or approved vector adapter
- Commercial use
- Commercial use permitted.
- Data boundary
- Embeddings may reveal source semantics; classify content before indexing.
Embedding bootstrapFirst 200M tokens free for several current models; not a monthly renewal.
WHY THIS MATTERS
One-time index-bootstrap allowance for supported embeddings and reranking. Covers the one-time work of turning a business's content into AI-searchable form, so building the initial semantic-search index does not carry a recurring fee.
Illustrative low-volume operation within the current documented allowance: A firm indexes 150 pages of service and policy content once at launch. That is a small fraction of the 200M-token one-time bootstrap allowance, so the initial index build costs nothing.
Expected light use: Initial embedding or reranking bootstrap for an approved content index.
- What consumes it
- Input tokens processed by eligible models; the account allowance is one-time rather than monthly.
- Fallback
- Workers AI embeddings or keyword search.
- Upgrade trigger
- Bootstrap allowance is consumed or model requirements change.
- Likely paid successor
- Voyage AI paid usage or approved embedding adapter
- Commercial use
- Commercial use subject to account terms.
- Data boundary
- Verify training and retention terms before private content use.
Realtime transcription20 realtime hours/month and two concurrent sessions.
WHY THIS MATTERS
Low-volume captions and voice input. Adds live captions or voice input where they genuinely help accessibility or usability, without a dedicated transcription contract.
Illustrative low-volume operation within the current documented allowance: A module offers voice-to-text intake for a handful of users a week, totaling three to four realtime hours a month—well inside the 20-hour/month allowance.
Expected light use: Occasional live captions or voice input, with limited batch transcription for supplied recordings.
- What consumes it
- Realtime minutes, batch minutes, and concurrent realtime sessions.
- Fallback
- Text input, uploaded transcript, or manual caption workflow.
- Upgrade trigger
- Hours or concurrency approach 75%, or batch processing is required.
- Likely paid successor
- Speechmatics paid plan or approved transcription adapter
- Commercial use
- Commercial use permitted subject to current terms.
- Data boundary
- Explicit microphone consent; no background recording; verify regional requirements.
Browser verification10 browser minutes/day on Workers Free.
WHY THIS MATTERS
Narrow critical-path browser monitoring. Automatically checks that the live site's most important pages and forms actually work, catching an outage before customers do.
Illustrative low-volume operation within the current documented allowance: A daily automated check loads the homepage, the contact form, and one service page—about ninety seconds of browser time a day, comfortably under the ten-minute daily allowance.
Expected light use: A narrow daily critical-path schedule covering a few important public pages and forms.
- What consumes it
- Browser session duration shared across Quick Actions and Browser Sessions.
- Fallback
- HTTP checks, manual QA checklist, and user-reported issue path.
- Upgrade trigger
- Critical-path schedule cannot fit within the daily minute budget.
- Likely paid successor
- Cloudflare paid browser usage or approved monitoring provider
- Commercial use
- Commercial use permitted.
- Data boundary
- Do not capture sensitive pages or secrets in screenshots.
Beta cited retrieval acceleratorOpen-beta allowance described as 20K queries/month, 100K files/instance, and 500 crawled pages/day; terms may change with notice.
WHY THIS MATTERS
Optional beta accelerator for natural-language retrieval and citations. An experimental way to let visitors ask natural-language questions and get cited answers pulled directly from the business's own content, without building a custom AI search pipeline from scratch.
Illustrative low-volume operation within the current documented allowance: A pilot deployment answers 500 natural-language questions a month over a 40-page content set—well inside the 20,000 monthly queries and 100,000-file beta allowance.
Expected light use: A controlled beta retrieval prototype over an approved website or document corpus.
- What consumes it
- Monthly queries, files per instance, and pages crawled; Workers AI and AI Gateway usage remains separate.
- Fallback
- Worker + D1/R2 + Vectorize + embedding adapter + conventional search.
- Upgrade trigger
- Beta terms, pricing, availability, or quality no longer meet requirements.
- Likely paid successor
- Provider-independent retrieval architecture
- Commercial use
- Approved limited while beta terms remain acceptable.
- Data boundary
- Never make the production promise depend on beta availability.
Verification dates describe the repository record, not a permanent guarantee. Re-verify every provider before proposal and quarterly during Care. Stale records must be quarantined rather than displayed as current.