Built for engineering teams
Your team spends $8k/mo
on AI. Where does it go?
CostLens gives you per-feature, per-developer cost attribution for OpenAI and Anthropic. Know exactly which prompts are burning your budget.
2 min setupClaude Code · Kiro · CursorFree for small teams
One line to track everything
Wrap your OpenAI or Anthropic client. Every request gets tagged with cost, model, latency, and feature attribution. No other code changes needed.
OpenAIAnthropicClaude CodeKiroCursor
app.ts
import { CostLens } from 'costlens' import OpenAI from 'openai' const costlens = new CostLens() const openai = costlens.wrapOpenAI(new OpenAI()) // Your code doesn't change. At all. const res = await openai.chat.completions.create({ model: 'gpt-4', messages: [{ role: 'user', content: 'Say hello' }] }) // → Routed to gpt-4.1-nano (10x cheaper) // → Quality preserved for this prompt
This is what you get in 2 minutes
Real dashboard. Real data. No setup wizards, no config files.
costlens.dev/dashboard
Team plan — costlens.dev/analytics
Upgrade for full analytics
Team plan unlocks cost by model, provider, team, and developer breakdowns.
Per-model cost breakdown
See exactly how much you spend on GPT-4o vs Claude Sonnet vs cheaper alternatives.
Savings tracked automatically
Every routed request shows what you would have paid vs what you actually paid.
See which feature burned $400 last Tuesday
Not averages. Not estimates. Exact cost, per prompt, per developer.
Attribution
Cost per feature, per developer
Tag prompts by feature, team, or user. Know that "code review" costs $340/mo and "customer support" costs $89/mo. Make informed decisions about where to optimize.
Safety
Kill switch for runaway agents
Burst detection catches infinite loops. One-click pause from Slack. Auto-resumes after 30 min.
Productivity
Prove AI makes your team faster
Track sessions, artifacts, commits, and time-to-ship. Weekly reports your VP can forward to the board.
Agent-native
Built into your coding agent
Works inside Claude Code, Kiro, and Cursor via MCP. Developers get cost visibility without leaving their editor. Zero context switching.
Optimization
Smart routing included
Automatically route simple prompts to cheaper models. Save 20-40% without changing your code.
Alerts
Budget limits with Slack alerts
Per-key daily budgets, burst rate limits, and spend velocity caps. Get alerted at 80% — or let the system auto-block before you overspend.
We built this because we got a $4,200 bill with zero explanation
Our team was using GPT-4 for everything — summarization, code review, customer support. Same model, wildly different value. CostLens started as an internal tool to figure out what was worth the spend and what wasn't.
2 min
to first cost report
20-40%
typical savings with routing
Pricing
Free for small teams. Pay when you need team features.
Free
$0
For individual developers.
Cost tracking
Smart routing
MCP Server (Claude Code, Kiro, Cursor)
OpenAI + Anthropic
1 GitHub repo
Burst protection (defaults)
Team
Popular$99/mo
For engineering teams.
Everything in Free
Full analytics dashboard
Response caching
2 scheduled reports (self only)
1 GitHub repo
Up to 5 team members
90-day data retention
Business
$299/mo
For scaling organizations.
Everything in Team
Unlimited reports + recipients
Unlimited GitHub repos
Unlimited team members
Unlimited data retention
Custom routing rules
Priority support
Stop guessing. Start measuring.
One npm install. First cost report in under 2 minutes. No credit card required.
Start free
npm install costlens