BazaarLinkBazaarLink
Sign in

Enterprise AI Spend Control Multi-layer Budgets · Circuit Breaker · Full Audit Trail

Every tier — org, team, member — has its own spend cap. Abnormal usage trips automatically. Every token is traceable.

Pay-as-you-go or enterprise monthly billing; zero model markup, 10% transaction fee at top-up

Contact Sales Get API Key
✕Switching models requires code changes
✓Zero switching cost, change one parameter to swap models
✕API key leaked or abused — only spotted on next invoice
✓Per-key daily / weekly / monthly cap; auto-disabled on overage
✕One AI bill across teams — no clear cost ownership
✓Allocate budget and usage reports by org, team, and member
✕A runaway script can burn $10k overnight
✓Circuit breaker cuts traffic via 1-min + 1-hr dual-window in real time
✕Devs sneak in expensive or un-vetted models
✓Allowed Models whitelist blocks them instantly

Quote First, Pay Later — Built for Corporate & Government Procurement

Many companies and government agencies need a formal quotation and invoice before payment can be approved internally. We fully support this workflow — just reach out.

1

Request a quotation

Email us your expected usage and requirements; we'll issue a formal quotation (TWD pricing available).

2

Receive the invoice for approval

We issue the invoice (Taiwan e-invoice) up front so your internal approval and procurement process can proceed.

3

Pay, then go live

Once payment arrives via bank transfer or monthly billing, your credit is activated immediately.

Email us for a quotation Available to companies, schools, and government agencies. Bank wire and Net-30 monthly billing (negotiable) supported.

Enterprise Usage Control

Real-Time Usage Reports

Spend and token dashboards broken down by model, team, member, and API key, with CSV, JSON, and print/PDF export for finance review.

Scoped Key Management

Create organization, team, member, and management keys with role-aware access. Enable, disable, rotate, and attribute usage without sharing one master key across teams.

API Key Spend Limits

Each API key supports independent daily/weekly/monthly spend limits — auto-disabled on overage, auto-resumes next period, with real-time usage tracking.

Monthly limit$750 / $1,000
$24
Now
$168
This week
$750
This month

API Key Management

Programmatic API key management under enterprise control — spend limits, expiry, and enable/disable per key.

sk-bl-prod-••••ProductionActive
sk-bl-dev-••••DevelopmentActive
sk-bl-ci-••••CI / CDDisabled

Three-Tier Budget Control

Set monthly budget caps independently at org, team, and member levels. Org-level uses SQL atomic deduction; team and member levels use Redis real-time tracking.

Org Budget
$50,000 / mo
Team Budget
$10,000 / mo
Member Budget
$2,000 / mo

Circuit Breaker (Spend Kill Switch)

Runaway AI scripts burning $10,000 overnight is no longer rare. BazaarLink's dual-window monitor (1-min + 1-hr) cuts upstream traffic the moment things go wrong — over 30× faster reaction than monthly caps.

  1. Every request
  2. Dual-window count
  3. Per-level thresholds
  4. 503 on breach
  5. Recovery after reset
Circuit breaker flow: a breach returns HTTP 503, and traffic recovers on its own once the window resets, with no manual intervention.

Dual-window counter

Per-minute and per-hour spend counters evaluated together; catches both bursts and slow leaks.

Org / Team / Member scopes

Each level has independent thresholds with auto fallback (member → team → org → default).

Redis real-time + 30s cache

Atomic check on every request with near-zero added latency.

Trip + auto recovery

Returns HTTP 503 immediately on breach; auto-recovers once the window resets on the clock boundary — no manual intervention needed.

Real scenario
A nightly batch job loops infinitely, firing 200 GPT-4 requests per minute → $5/min threshold trips → automatic HTTP 503 rejects further calls → auto-recovers after cooldown, total damage < $5.

Allowed Models Whitelist

Restrict your Org / Teams to approved models only. Compliance, cost, and quality solved in one place — no need to hard-code blocklists.

  • Stop devs from sneaking expensive models (e.g. Claude Opus) into production
  • Block providers that haven't passed legal review to prevent data exfiltration risk
  • Add / remove models without redeploying — takes effect instantly
  • Returns HTTP 403 with a clear error message when a request hits a blocked model
  • Audit log captures who changed the list and when

Organization & Team Management

Multi-org architecture with four RBAC roles (admin, billing viewer, team admin, member), domain whitelisting, invite links, cost centers, and team/member attribution.

Org AdminBilling ViewerTeam AdminMember

Flexible Billing

Pay As You Go

Prepaid credits for developers and smaller teams that want immediate access without a sales process.

Monthly Enterprise Billing

Monthly billing with credit limits, payment terms, billing contacts, and invoice records for larger organizations.

Pricing standard

Based on each provider's official list price with zero model markup; the platform fee is charged at top-up (deposit): a 10% transaction fee (credit-card top-ups add a US$0.60 flat fee). Currently offered payment methods, applicable Taiwan VAT, and invoice eligibility are shown before payment. Enterprises can arrange monthly billing, Net-30 (negotiable).

View live per-model pricing in TWD View full billing details

Security & compliance summary

Conversation content not retained at the relay layer; only usage metadata kept
Encrypted connections, hashed customer API keys, access control and audit trails
Security incidents notified promptly after confirmation (72-hour target), per PDPA
Allowed-models whitelist to restrict providers per your data policy

Security overview and other compliance documents are available on request.

Zero Retention
prompt
→
→
model
BazaarLink stores nothing
Billing metadata only
Never used for training

Zero Data Retention

BazaarLink itself does not store the content of your prompts or responses — only token counts, timestamps, and billing metadata.

Note: Upstream model providers follow their own data-handling policies. BazaarLink cannot guarantee provider-side behavior on their behalf.

Frequently asked questions

Can you issue a quotation and invoice first, with payment afterwards?+
Yes. Many companies and government agencies need a formal quotation and invoice before payment can be approved internally — we fully support this workflow. Email sales@bazaarlink.ai with your expected usage; we'll issue a formal quotation and the invoice up front, and activate your credit as soon as payment arrives via bank transfer or monthly billing.
Does BazaarLink support Taiwan e-invoice (統一發票)?+
Yes. BazaarLink is operated by 集聯科技有限公司 (Tax ID 62232477, registered in Toufen City, Miaoli, Taiwan), and issues Taiwan e-invoices that meet public-sector and enterprise accounting requirements.
Which AI models does BazaarLink support?+
BazaarLink integrates 170+ AI models, including OpenAI (GPT-4o, GPT-4 series), Anthropic (Claude series), Google (Gemini), Meta (Llama), and other major providers. Enterprise Plan customers can further restrict access via the Allowed Models allowlist.

Ready to transform how you build with AI?

Join the companies already building with BazaarLink

Contact Sales Get API Key
Support
Support
Hi! How can we help you?
Send a message and we'll get back to you soon.