Monitor Anthropic costsper feature.
Route your Anthropic client through Spillr and instantly see how much each feature costs in Claude tokens — broken down by support-bot, summariser, or any label you choose.
Supports all Claude models · No prompt storage · 2-minute setup
import Anthropic from '@anthropic-ai/sdk';
const client = new Anthropic({
baseURL: 'https://proxy.spillr.io/anthropic', // ← one line change
defaultHeaders: {
'X-Spillr-Key': process.env.SPILLR_KEY,
'X-Spillr-Feature': 'support-bot', // ← tag per feature
},
});
// Your existing Anthropic code — unchanged.
const res = await client.messages.create({
model: 'claude-haiku-4-5',
max_tokens: 512,
messages: [{ role: 'user', content: prompt }],
});The problem
Claude tokens are expensive. Do you know which feature is burning them?
Claude Opus is among the most capable — and most expensive — models available. When you ship multiple AI features against the same API key, Anthropic's usage page cannot tell you which one is responsible for a spike. Spillr adds that layer of visibility in two minutes.
Aggregated billing only
Anthropic groups usage by API key, not by feature. A costly feature hides inside the total.
No automatic guardrails
There is no native way to block Claude requests when a feature exceeds a spend threshold.
Input + output token asymmetry
Claude pricing differs for input vs. output tokens. Spillr tracks both and calculates exact cost per call.
Features
Full cost visibility for Anthropic workloads.
Per-feature cost breakdown
Tag calls with X-Spillr-Feature and see Claude costs grouped by feature in the dashboard — updated in real time.
Proactive budget alerts
Set a monthly or daily limit per feature. Spillr emails you when spending hits your alert threshold.
Hard spend limits
When a budget is exhausted, Spillr returns a 429 before the request reaches Anthropic. Your bill stops at the cap.
Streaming support
Anthropic SSE streaming passes through unchanged. Usage data is captured from the final stream event.
Privacy by design
Prompts and completions are never logged. Spillr captures only token counts and computed cost.
Multi-provider dashboard
Anthropic, OpenAI, and Google Gemini in one view. Set cross-provider budgets with the same interface.
FAQ
Common questions about Anthropic cost monitoring.
Does Spillr store my Claude prompts or responses?
No. Spillr records only token counts (input and output), the model name, and the calculated cost. Your messages and completions are never stored — they pass straight through to Anthropic.
Which Claude models does Spillr support?
All Claude models are supported: Claude Opus 4, Claude Sonnet 4, Claude Haiku 4, Claude 3.5, and any future models Anthropic releases. Pricing is looked up automatically from the model identifier in each request.
Does Anthropic streaming work through Spillr?
Yes. Anthropic SSE streaming passes through unchanged. Token counts from the message_delta event are captured and logged after the stream completes.
How is this different from Anthropic's built-in usage page?
Anthropic's dashboard shows aggregate token usage per API key. Spillr adds the per-feature dimension — you can see that your support-bot feature costs $20/month while your summariser costs $3.
Can I monitor both Anthropic and OpenAI from one dashboard?
Yes. Spillr supports OpenAI, Anthropic, and Google Gemini from one proxy. All costs appear in a single dashboard with consistent budgets and alerts across providers.
Know your Claude costs in 2 minutes.
Change one line. Add two headers. See per-feature Anthropic costs right away.
Create free account →