Decouple pricing
from code.

const tokens = await anthropic.count({
  model: "claude-sonnet-4-6",
  input: req.body.prompt,
  output: res.body.completion,
});

const cost = tokens.input * 0.003
           + tokens.output * 0.015;

const customer = await db.query(
  'SELECT plan, balance FROM customers'
  + ' WHERE api_key = $1', [req.headers.key]
);

await stripe.invoiceItems.create({
  customer: customer.stripe_id,
  amount: Math.ceil(cost * 100),
  currency: 'usd',
});// Latency: 250ms. High risk of inference timeout.

plugins:
  - name: aforo-metering
    config:
      architecture: async_sidecar
      product_type: AGENTIC_API
      metrics:
        - input_tokens
        - output_tokens
        - vector_db_lookups
      pricing_model: compound_mn
      margin_guard:
        l1_warn:     20%
        l2_throttle: 10%
        l3_block:     0%
      enforcement:   redis_fail_open
      inference_latency_added: 0ms// Inference Latency: 0ms. Async sidecar. Out-of-band state push.