Documentation

1. Create an Account

Sign up for TensorCortex and get your API key instantly. No credit card required for the free tier with 10,000 requests per month.

2. Use Your Existing SDK

No new SDK needed! Just change your base URL. Works with OpenAI, Anthropic, and all provider SDKs you already use.

# Just change the base URL - that's it!
export OPENAI_BASE_URL=https://openai.tensor.cx/v1
export OPENAI_API_KEY=your-openai-api-key

⚠️ Important: Include /v1 in your base URL. TensorCortex is a pure pass-through proxy.

3. Make Your First Request

Your existing code works instantly. Zero code changes required - just point to TensorCortex.

from openai import OpenAI

# Just change base_url - your code stays the same!
client = OpenAI(
    api_key="your-openai-api-key",
    base_url="https://openai.tensor.cx/v1"
)

response = client.chat.completions.create(
    model="gpt-4o",
    messages=[{"role": "user", "content": "Hello!"}]
)

⚠️ BYOK (Bring Your Own Keys): Use your actual provider API key. TensorCortex uses your OpenAI key to call OpenAI directly - zero markup.

4. Enjoy Automatic Benefits

That's it! You now get automatic caching (60% cost savings), global edge routing, and real-time cost tracking - all with zero configuration.

Automatic Caching

TensorCortex automatically caches deterministic requests (temperature=0) to save costs and reduce latency. Cache hits are served from global edge storage in under 50ms.

Provider Routing via Subdomain

Route to different providers by changing the subdomain. Models use their native names.

Cortex Configuration (Dashboard)

Sign up for the dashboard at app.tensorcortex.com to configure advanced features per Cortex:

Use Fallbacks for Production

Always configure at least 2-3 fallback models for production applications. This ensures your app stays online even during provider outages.

Monitor Your Usage

Use the dashboard analytics to track costs across providers. Identify opportunities to optimize by switching to more cost-effective models for certain use cases.

Cache Responses When Possible

For deterministic queries (temperature=0), enable response caching to reduce costs and latency for repeated requests.

Use Streaming for Long Responses

Enable streaming for chat applications to provide a better user experience with immediate feedback as tokens are generated.

API Key Security

Your API keys provide full access to TensorCortex. Keep them secure and never expose them in client-side code.

Data Privacy

TensorCortex processes requests at the edge and does not store your prompts or responses. Your provider API keys are encrypted with AES-256-GCM. Enterprise customers can request custom data residency options.