$188.43
Highshopping/catalog.py::recommend_products
gpt-4o · Test Gemini Flash
AI spend visibility for engineering teams
Find AI cost leaks.Without rewriting your application.
Keep your standard OpenAI and Anthropic clients. Add one decorator to categorise supported AI workflows, then see where spend, token use, and cache efficiency need attention.
No proxy in your LLM request path. Your OpenAI and Anthropic calls go directly to the provider.
See where AI money goes
Find missed savings fast
Control costs before they scale
Automatic instrumentation for supported Python OpenAI and Anthropic SDK operations.
Supports OpenAI and Anthropic tracking today. Includes cache efficiency reporting when providers return cache-token usage.
Yesterday spend
$421.80
Potential savings
$4,812/mo
Projected monthly $12,654
Calls 18,291
3 recommendations
$94.12
Mediumshopping/reviews.py::summarize_reviews
claude-sonnet-4 · Compare Qwen or Mistral
$51.06
Lowshopping/search.py::extract_filters
gpt-4o · Review DeepSeek or Llama
What SpendLens helps you answer
Know what is costing money, and what to review first.
Most teams can see a total AI bill. SpendLens helps explain where that bill came from and which savings opportunities are worth testing.
Open Model SavingsWhich project is spending most?
Break down AI cost by project, model, provider, and workload.
Which workload is expensive?
Find the product flows or endpoints creating the biggest AI bill.
Where can we save?
Review cheaper model options, cache gaps, and prompt waste signals.
What should we test first?
Prioritize savings by estimated impact, confidence, and risk.
How it works
From AI calls to cost clarity.
01
Set up SpendLens
Install the Python SDK and add your SpendLens project and API key as environment variables. Node.js SDK coming soon.
02
Existing code? Add a decorator
Add @spendlensai.observe to an AI workflow. Keep your existing OpenAI or Anthropic client unchanged.
03
Need precision? Tag the client
Use track() and client.tag() when you want exact task, feature, experiment, or endpoint attribution.
04
See what is costing money
Open the dashboard to compare workloads, models, token usage, cache efficiency, and savings opportunities.
Savings insights
Find the AI spend you should review first.
SpendLens does not just show charts. It points to practical savings opportunities such as expensive model choices, poor cache usage, oversized prompts, and high-token workloads.
Expensive model choices
See when simple workloads may be running on unnecessarily expensive models.
Prompt and token waste
Find large templates, excessive context, repeated instructions, and long outputs that increase spend.
Savings worth testing
Review estimated savings, confidence, and risk before making production changes.
Executive-ready reporting
A simple daily summary of spend and savings.
Get a compact report showing yesterday's spend, the top cost driver, and the top recommendation to review.
View sample reportYesterday
$421.80
Projected
$12,654
Savings
$4,812/mo
Recommendation confidence
HighTest `shopping/catalog.py::recommend_products` on Gemini Flash or Qwen Flash.
SpendLens found a high-cost workload, identified a cheaper model option worth testing, and estimated $4,812/month in potential savings.
High
Strong workload match and low migration risk
Medium
Worth testing with sample replay
Low
Review manually before migration
Privacy-aware tracking
Cost visibility without storing full prompts by default.
By default, SpendLens captures operational metadata and reusable system/developer templates—not user prompts or model responses.
User prompts and model responses are not stored by default
Template-only prompt sampling is the default
Metadata-only mode is available
API keys are stored as hashes only
Detected workload types can be corrected
Pricing
Start free. Prove the savings. Then scale.
Annual plans include 2 months free.
Free
$0/month
1,000 events/month, 1 project, test API key, basic dashboard
Start freeStarter
$49/month
100,000 events/month, 1 project, 14-day retention
Pro
Recommended$199/month
1,000,000 events/month, 5 projects, 90-day retention
Growth
$499/month
5,000,000 events/month, 20 projects, priority support
FAQ
What teams ask before installing.
Do you proxy my LLM traffic?
No. Your OpenAI and Anthropic calls continue going directly to the provider. The SpendLens SDK sends usage metadata after calls complete.
What do I get after installing?
A dashboard showing which projects, workloads, models, and providers are driving AI spend, plus recommendations for savings to review.
Do you store prompts?
No full user prompts are sent by default. The default template-only mode may use system prompt or template text for workload classification.
Which providers are supported?
SpendLens currently tracks OpenAI and Anthropic usage. Recommendations can compare against OpenAI, Claude, and other leading model providers.
Are savings guaranteed?
No. Savings are estimates. SpendLens helps you find opportunities worth testing before changing production workloads.
Do I need to tag every task?
No. Add @spendlensai.observe to a business workflow. Use client tags only when you need more precise per-call attribution.
See where your AI money is going this week.
Install the SDK, send usage metadata, and get a dashboard that shows spend, cost drivers, and savings opportunities.