thealpha.ai logo
See your agent costs — free
ProductArenaPricingDocsBlogSee your agent costs — free

Docs · Quickstart

Point your baseURL at Alpha.

One line changes. Nothing else does. Every agent call runs through the operating layer — first controlled run in under two minutes. No SDK, no re-architecture, no excuses.

  1. 1 · Swap your baseURL

    Alpha is OpenAI-compatible (and Anthropic-compatible). Keep your SDK — point it at the gateway, authenticate with your Alpha key, and prefix models as provider:model.

    from openai import OpenAI

client = OpenAI(
    base_url="https://api.thealpha.ai/v1",       # ← the only change
    api_key="ALPHA_GATEWAY_KEY",  # your Alpha key (gw_…), not a provider key
)

resp = client.chat.completions.create(
    model="openai:gpt-4o-mini",   # provider:model
    messages=[{"role": "user", "content": "ping"}],
)

  2. 2 · Add your provider keys once

    BYOK: add your OpenAI, Anthropic, Google, Bedrock, or other provider keys in the dashboard. They're stored server-side, isolated per workspace, and scoped per agent — Alpha routes through your own accounts and never marks up model costs.

  3. 3 · Watch it in the dashboard

    Your first call shows up immediately: cost, tokens, latency, which agent, which budget. That's the run you couldn't see before.

Want to see the savings before you switch? Run Arena — free.