Private AI Gateway

One endpoint, ranked models, keys with hard limits.
https://omnitrix.space

01 Claude Code

Add to ~/.bashrc (or ~/.zshrc), restart terminal, run claude.

# Omnitrix gateway for Claude Code
export ANTHROPIC_BASE_URL="https://omnitrix.space"
export ANTHROPIC_AUTH_TOKEN="sk-omni_YOUR_KEY"
export ANTHROPIC_MODEL="newcomer"
export ANTHROPIC_DEFAULT_SONNET_MODEL="newcomer"
export ANTHROPIC_DEFAULT_OPUS_MODEL="active"
export ANTHROPIC_DEFAULT_HAIKU_MODEL="newcomer"

02 OpenAI-compatible (Python)

Works with the official openai SDK and anything that speaks the OpenAI chat API.

from openai import OpenAI

client = OpenAI(
    base_url="https://omnitrix.space/v1",
    api_key="sk-omni_YOUR_KEY",
)

resp = client.chat.completions.create(
    model="newcomer",
    messages=[{"role": "user", "content": "Hello from Omnitrix"}],
    stream=True,
)
for chunk in resp:
    print(chunk.choices[0].delta.content or "", end="")

03 curl — 10-second smoke test

curl -s https://omnitrix.space/v1/chat/completions \
  -H "Authorization: Bearer sk-omni_YOUR_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"newcomer","max_tokens":32,
       "messages":[{"role":"user","content":"Reply with OK"}]}'

04 Model ranks

ModelWhat it isSpeedUse for
newcomerDeepSeek-V4.1-Flash · GLM-5.3-Flash (RiverAI pool)~1–2sDaily coding & chat — the default
activegrok-4.7 family (DGB pool)10–25sHeavy reasoning fallback
progrok-imagine-image · -quality~5sImage generation
legendgrok-imagine-video + flagship modelsmin.Video / top-tier (rank-gated)

Fallback is automatic: if the fast tier fails, requests reroute to active. Your key's rank decides which models you may call.

05 Image generation

curl -s https://omnitrix.space/v1/chat/completions \
  -H "Authorization: Bearer sk-omni_YOUR_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"pro","max_tokens":200,
       "messages":[{"role":"user","content":"a red circle on white background"}]}'
# → response contains a hosted image URL