Next.js starter app
Dashboard, routes, setup docs, environment validation, and a focused demo workflow.
For paid AI products
Track AI costs by workflow, not just tokens.
A source-code Next.js starter for paid AI apps. Add an internal usage layer, workflow-level cost tracking, soft caps, billing snapshots, and admin visibility before margins break.
Source-code starter
Outcome cost
Retry spend
Soft caps
Billing snapshots
Cost / report
$0.21
profitable
PDF Q&A
$0.48
context spike
Retry spend
31%
needs review
Workflow
Outcome
Cost
Cap state
report_generation
usable
$0.21
ok
pdf_question
usable
$0.48
soft cap
ticket_reply
failed retry
$0.16
leaking
Every model call is tagged before it runs.
customer, workflow, model, entitlement snapshot, retries, final outcome
What you get
A starter repo for the AI usage layer most teams build too late.
The first version is designed to be practical source code, not a separate SaaS service. Drop it into a paid AI app and route model calls through one controlled boundary.
Internal AI usage layer
One typed client for model calls, provider routing, retries, caps, and outcome logging.
AI usage ledger
Self-contained records for customer, workflow, cost, price ID, entitlement, and final outcome.
Admin dashboard
Workflow cost, retry spend, failed runs, outlier users, plan usage, and margin leakage.
Billing connection
Plan snapshots, soft caps, upgrade prompts, and reconciliation-friendly event records.
Docs and support boundaries
Setup guide, architecture notes, buyer license, and clear limits for included support.
The real failure mode
AI billing fails quietly before it fails loudly.
Most teams notice the problem after launch: outlier users, retries, context spikes, failed runs, and billing records that do not explain what happened.
Tokens do not map to pricing
Token counts explain provider cost. They do not tell you whether the user received a valuable outcome.
Retries hide margin leaks
Agentic workflows can spend heavily on failed or retried runs before anyone checks the dashboard.
Usage and billing drift apart
If plan state is looked up later, upgrades and changing limits make reconciliation unreliable.
The V1 system
One internal usage layer for every AI workflow.
Start inside the app, but keep the boundary shaped like a gateway so it can become a relay later.
Tag
customer, workflow, environment, price ID, entitlement snapshot
Cap
soft limits, context windows, model allowlists, hard risk blocks
Route
provider keys stay behind one internal client
Log
tokens, cost, latency, retries, failures, final outcome
Price
see workflow cost, usage by plan, and quiet margin leaks
Developer contract
Keep product code clean. Put AI accounting behind one call.
The template should feel like a normal Next.js app, but every provider call must pass through a typed usage boundary that records cost and outcome.
await aiUsage.run({
customerId,
userId,
workflow: "pdf_question",
feature: "rag",
environment: "production",
entitlement: planSnapshot,
caps: {
maxContextTokens: 12000,
softDailyCost: 4.00
},
action: async ({ model }) => {
return streamAnswer({ model, question, chunks })
}
})
ai_usage_events
self-contained billing record
customer_id
who paid
workflow_id
what outcome
provider_cost
what it cost
retry_count
where margin leaks
active_price_id
billing snapshot
usable_result
did user get value?
How it works
Wrap the AI call. Log the outcome. Price the workflow.
The buyer should not have to redesign billing architecture from scratch. The template gives them a working pattern they can adapt.
Wrap AI workflows
Call aiUsage.run() from product workflows instead of calling providers directly.
Record cost and outcome
Log tokens, retries, provider cost, failed runs, usable result, and plan snapshot.
Control margin
Use soft caps, workflow dashboards, and billing snapshots to catch margin leaks early.
Use cases
Built for paid AI workflows where usage can get expensive.
Track retrieval, context windows, file caps, and cost per answered question.
Measure resolved tickets, retries, failed runs, fallback spend, and plan limits.
Price reports by completed output while tracking model mix and retry leakage internally.
Reuse a client-ready pattern for billing, caps, admin visibility, and handoff docs.
Product principle
Users buy outcomes. Owners track economics.
Keep the customer-facing plan simple: reports, documents, tickets, runs. Internally, track the cost and margin behind every unit.
User sees
Owner tracks
Decision
100 reports / month
cost per report
keep
500 document questions
retrieval + context cost
watch
1,000 resolved tickets
retry + failure share
reprice
50 agent runs / day
tool calls + fallback cost
soft cap
What ships in V1
The pieces that protect margin and support billing audits.
AI usage ledger
Every event stores customer, workflow, model, tokens, cost, retries, outcome, price ID, and entitlement snapshot.
Workflow profitability
Dashboard views by workflow, model, customer, retry spend, failed runs, and plan usage.
Soft caps first
Warnings and upgrade prompts before hard blocks, unless cost risk is genuinely dangerous.
Privacy-safe defaults
Prompt and body logging stay off by default. Cost metadata remains useful without sensitive content.
Billing snapshots
Events are self-contained enough to re-derive charges after plan changes and mid-cycle upgrades.
Gateway-shaped boundary
One internal AI client today, clean extraction path to relay/gateway later.
Planned pricing
Expected one-time pricing after beta.
Joining early access is free and does not create a purchase or reserve a price. These ranges test pricing expectations for the future source-code release.
Founding Early Access
Expected $49Planned first paid release for early builders who help shape V1.
- Paid access opens with beta
- V1 codebase when released
- Setup guide
- Bugfix updates
Pro
$149-$199For indie hackers building paid AI apps.
- Full template
- Admin dashboard
- RAG/caps example
- Priority issue support
Agency
$299-$499For client projects and repeat delivery.
- Client-use license
- White-label rights
- Handoff docs
- Priority support
FAQ
Positioned for paid AI apps, not generic SaaS.
Is this another SaaS boilerplate?
No. It includes SaaS foundations where needed, but the product focuses on AI usage accounting, cost caps, billing snapshots, and workflow profitability.
Is this a separate gateway service?
Not in V1. It starts as app-level middleware with a gateway-shaped boundary so it can be extracted later.
Do users see tokens and model costs?
No. Users see simple units like reports, documents, tickets, or runs. Owners see internal cost and margin by workflow.
Do you log prompts?
Prompt/body logging is off by default. The ledger records metadata for cost, reliability, caps, and billing audits.
Early access
Building a paid AI app?
Join the early list if you want the first working demo, beta access, or want to help shape the usage layer before V1 ships.