New 5-layer detection engine

Your AI budget has a hidden leak

Most teams unknowingly pay for the same AI call hundreds of times. TokenLens reads your logs and shows you exactly where the money goes.

Upload logs free → How it works

3 free analyses · No signup · No code changes · Results in 30 seconds

Savings Found
Analyzing
$0.00
wasted on duplicate calls this month
Total API spend $4,218.00
Duplicate calls found 2,847
Recoverable waste 84.1%
After optimization $669/mo
gpt-4o
Switch to mini → save 80%
Cache hit
999× identical prompt
84%
average waste detected
5
detection layers
<10s
for 10K records
$0
code changes needed
How it works

Three steps. Thirty seconds.

No agents, no SDKs, no code changes. Just upload a log file and see your waste.

📁

Upload your logs

Drop any AI API log — JSON, JSONL, or CSV. We auto-detect OpenAI, Anthropic, Datadog, and custom formats.

🔍

We find the waste

Our 5-layer engine catches exact duplicates, near-matches, templates, and paraphrased calls bleeding your budget.

💰

Fix & save

Get specific recommendations: which prompts to cache, which models to downgrade, and who's overspending.

Detection engine

Five layers deep. Nothing escapes.

Each layer catches what the previous one misses, with increasing sophistication.

1

SHA-256 Exact Match

Identical prompts, byte-for-byte

100% precision
2

MinHash LSH

Whitespace & formatting variants

<2% false positive
3

N-gram Overlap

Reordered words, same meaning

95%+ recall
4

Template Detection

Same prompt, different variables — "Summarize {X}" called 500×

pattern match
5

Semantic Similarity

Paraphrased intent — different words, same question

TF-IDF cosine
Who it's for

Built for teams spending $1K+/mo on AI APIs

🏗️

Engineering teams

Find which microservices make redundant LLM calls and where a simple cache would save thousands.

💻

Copilot & Cursor users

Prove ROI on your AI coding tools. Find ghost seats, duplicate completions, and optimize token limits. Learn more →

📊

Platform teams

See which teams and services are the biggest spenders and where model downgrades won't impact quality.

💼

Engineering managers

Get weekly reports showing AI waste trends, projected savings, and actionable fixes with dollar amounts.

Pricing

Stop overpaying for AI. Start with a free scan.

Every paid tier delivers minimum 3× ROI on your AI spend.

Free
$0
forever
"See what you're wasting"
  • 3 free analyses — no signup
  • 10 uploads/month with account
  • Up to 5K records/file
  • 3-layer detection
  • Copilot & Amazon Q basic analysis
  • Last 5 scans saved (7 days)
  • CSV export & cluster feedback
Get started free
Starter
$99
per month
"Fix waste systematically"
Typical ROI: 3–5× on $2K+/mo AI spend
  • 25 uploads/month, 50K records
  • Full 5-layer detection
  • 5 team members & shared scans
  • Copilot/Cursor/Amazon Q deep analysis
  • Caching & model downgrade analysis
  • Waste trend charts (30 days)
  • Adaptive thresholds & pattern library
  • Email alerts (5/day)
  • PDF, JSON & CSV reports
  • 90-day scan history
Start free trial →
Growth
$299
per month
"Manage spend across teams"
Typical ROI: 10× on $15K+/mo AI spend
  • Unlimited uploads, 500K records
  • 20 seats & team management
  • VS Code extension & cloud sync
  • Anomaly detection & prompt scoring
  • Token efficiency & cost forecasting
  • Multi-model price simulator
  • ROI report generator
  • Datadog & Splunk connectors
  • Slack, webhook & budget alerts
  • Cross-tool comparison (Copilot vs Cursor)
  • 1-year scan history
Start free trial →
Enterprise
Custom
contact sales
"Govern AI spend org-wide"
  • Everything in Growth
  • Unlimited seats & records
  • Team management & RBAC
  • SSO (SAML/OIDC)
  • Activity audit log
  • Self-hosted deployment
  • Custom integrations
  • Dedicated CSM
  • SLA guarantee
  • SOC 2 & HIPAA ready
Contact sales →
FAQ

Common questions

Do you need access to my API keys?+
No. TokenLens is completely read-only. We analyze log files you upload — we never make API calls on your behalf and never touch your production systems. Growth plans can pull usage data directly via Datadog and Splunk connectors.
What log formats do you support?+
We auto-detect logs from OpenAI, Anthropic, Google, Grok/xAI, GitHub Copilot, Amazon Q, Cursor, Codeium, Tabnine, LiteLLM, Helicone, Langfuse, Datadog, Splunk, and any JSON/JSONL/CSV with prompt and token fields. We also accept TokenLens VS Code extension exports. Our multi-layer format detection handles most custom formats automatically.
How accurate is the waste detection?+
Our SHA-256 layer has 100% precision (zero false positives). The full 5-layer pipeline achieves over 95% recall on exact duplicates and over 70% on paraphrased prompts with under 2% false positive rate. The learning engine continuously improves accuracy based on your feedback.
What's the difference between Starter and Growth?+
Starter ($99/mo) is for individual devs and small teams — you get full 5-layer detection, 5 team members with shared scans, caching analysis, model optimizer, Copilot/Cursor/Amazon Q deep analysis, waste trend charts, adaptive thresholds, pattern library, and PDF reports. Growth ($299/mo) adds 20 seats with full RBAC (admin/member/viewer), VS Code extension with cloud sync, anomaly detection, prompt scoring, token efficiency, cost forecasting, multi-model price simulator, ROI reports, cross-tool comparison, Datadog/Splunk connectors, Slack/webhook alerts, and budget monitoring.
Is my data safe?+
Log data is processed in memory and encrypted at rest (Starter+). Free tier data is retained for 7 days. Enterprise customers can self-host on their own infrastructure with SOC 2 and HIPAA compliance options.
What if I'm not sure my team has duplicates?+
That's the point — most teams don't know. In testing with production-like data, the average waste rate is over 80%. Upload a week of logs and see — you get 3 free analyses without even signing up.

Stop paying for the same answer twice

Upload your first log file in 30 seconds. No signup required.

Try TokenLens free →