delta_x generates the response. delta_y checks it every time. Drop-in replacement for OpenAI and Anthropic APIs.
# Before
client = OpenAI(api_key="sk-...")
# After — same code, smarter output
client = OpenAI(
api_key="dk-...",
base_url="https://deltaapi.duckdns.org/v1"
)
response = client.chat.completions.create(
model="delta_x",
messages=[{"role": "user", "content": "Hello!"}]
)# 1. Set environment variables
export ANTHROPIC_BASE_URL="https://deltaapi.duckdns.org"
export ANTHROPIC_API_KEY="dk-..."
# 2. Run Claude Code — that's it
claude
# Windows (PowerShell)
$env:ANTHROPIC_BASE_URL="https://deltaapi.duckdns.org"
$env:ANTHROPIC_API_KEY="dk-..."
claudeClaude Code will use delta_x as the model with delta_y reviewing every response in the background.
Every response goes through a two-model pipeline
Your prompt hits delta_x, a 120B MoE model. It produces a fast, high-quality response.
Simultaneously, delta_y checks the response for errors, bugs, and missing information.
If delta_y finds issues, delta_x automatically revises. You get the corrected answer.
Works with any SDK. Change base_url and api_key — done. Claude Code, Cursor, Copilot all supported.
Send images — delta_y analyzes them and adds context before delta_x responds.
Full SSE streaming. Responses start appearing in under 2 seconds on average.
No subscriptions. Top up credits and use them. Unused credits never expire.
Pay only for what you use. No monthly fees.
$1.00 / 1M input tokens · $1.00 / 1M output tokens · Secured by Paddle