1. Connect your code — 2 lines

Replace your OpenAI key (or any provider key) with your SmashCost key. Works with any OpenAI-compatible SDK.

client = OpenAI(
    base_url="https://api.smashcost.com/v1",
    api_key="sk-smashcost-..."  # your SmashCost key from the form above
)

Copy

That's it. SmashCost automatically routes to the best model for each task. The model field in your request is the base preference — SmashCost may route to another available model. The model field in the response shows which model actually ran.

# One-off override — this single request uses gpt-4o exactly
response = client.chat.completions.create(
    model="gpt-4o",
    messages=[...],
    extra_headers={"X-Model-Override": "gpt-4o"}
)

# The model field in the response confirms what actually ran:
# response.model  →  "gpt-4o-mini"  (what SmashCost chose)
# response.model  →  "gpt-4o"       (if you used X-Model-Override)

# 1. Free — the OpenAI `user` field, if you already send it, is used
#    as customer attribution automatically. No code change needed.
response = client.chat.completions.create(
    model="gpt-4o",
    messages=[...],
    user="acme_corp"
)

# 2. Set-and-forget — a key's "Default tags" (Access tab, below) label
#    ALL of that key's traffic, e.g. a key named "chat-widget" tagged
#    with default feature "chat-widget".

# 3. Most precise — per-request headers override both of the above
response = client.chat.completions.create(
    model="gpt-4o",
    messages=[...],
    extra_headers={
        "X-SmashCost-Tag-Feature": "reports",
        "X-SmashCost-Tag-Customer": "acme_corp",
        "X-SmashCost-Tag-Environment": "production"
    }
)

Select a routing mode above to see the trade-off before activating.

Alert threshold ($/hour)

Email address