Track per-feature AI cost with cost tags

The most common surprise on an LLM bill is finding out one feature you barely thought about — a background summarizer, a moderator agent, a debug helper — is silently 40% of your spend.

Cost tags fix that with one header. Tag every request with the feature it powers and we aggregate per tag for you.

Tag a request

Add an x-aig-tag header. Free-form string. Use whatever taxonomy you like.

curl https://api.aigateway.sh/v1/chat/completions \
  -H "Authorization: Bearer sk-aig-..." \
  -H "x-aig-tag: summarize" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "moonshot/kimi-k2.6",
    "messages": [...]
  }'

From the OpenAI SDK

Most SDKs let you pass extra headers per request.

from openai import OpenAI

client = OpenAI(
    base_url="https://api.aigateway.sh/v1",
    api_key="sk-aig-...",
)

resp = client.chat.completions.create(
    model="anthropic/claude-sonnet-4.6",
    messages=[...],
    extra_headers={"x-aig-tag": "summarize"},
)

Read spend per tag

GET the usage endpoint with a window. Optionally filter by a single tag.

curl https://api.aigateway.sh/v1/usage/by-tag?month=2026-04 \
  -H "Authorization: Bearer sk-aig-..."

# => [
#   { "tag": "summarize", "requests": 8210, "cost_cents": 4210 },
#   { "tag": "chat",      "requests": 3104, "cost_cents": 9830 },
#   { "tag": "rerank",    "requests":  912, "cost_cents":  118 }
# ]

Hard caps per tag

If a feature must not exceed a budget — say a free-tier user-facing chat — set a hard cap. We enforce it before dispatch, so you never get charged for a request that was going to be over-budget.

curl -X POST https://api.aigateway.sh/v1/budgets \
  -H "Authorization: Bearer sk-aig-..." \
  -d '{ "tag": "summarize", "monthly_cap_cents": 50000 }'

Related guides