OPENAI-COMPATIBLE SETUP

Connect once.
Use Binah everywhere.

Set one HTTPS base URL and one operator-issued API key in your terminal, application, notebook, API client, or server-side IDE workflow.

Start setup Choose a model

AUTHENTICATIONBearer API key

Issued by a Binah operator during controlled launch.

MODEL SELECTIONExact IDs only

Use a callable ID from the live model catalogue.

API SURFACEChat completions

Streaming and non-streaming requests are supported.

QUICKSTART

From key to first response.

No public checkout is active. Start with the key your Binah operator assigned to your tenant.

1
Get an operator-issued key
Sign in to the customer portal if your account has been provisioned, or contact your Binah operator. Keys are tenant-isolated and shown only when created.
Open customer portal
2
Keep the key local
Store it in your shell, password manager, deployment secret, or server environment. Never commit it to source control or expose it in browser code.
3
Call an exact model
No exact model is currently callable. Restore an admitted endpoint before running the placeholder examples below.
Browse callable models

REAL-TIME INFERENCE

Run real-time inference.

Send your first OpenAI-compatible request and return a model response immediately.Real-time inference is the best fit when a person or application is waiting for the answer. Existing OpenAI clients only need a Binah API key, the Binah base URL, and an exact model ID.

Request-response by default. Use non-streaming for one complete answer or enable streaming for progressive output.
Real gateway execution. The secure playground spends customer credit and returns the provider response—never a simulated demo.
OpenAI-compatible shape. Read the assistant text from choices[0].message.content.

Open secure playground View client examples

LIVE API CONTRACT<callable-model-id>

REQUEST

POST /v1/chat/completions
{
  "model": "<callable-model-id>",
  "messages": [
    { "role": "user", "content": "Summarize this risk in three bullets." }
  ],
  "stream": false
}

RESPONSE SHAPE

{
  "id": "chatcmpl_…",
  "model": "<callable-model-id>",
  "choices": [{
    "message": {
      "role": "assistant",
      "content": "<model response>"
    }
  }],
  "usage": { "total_tokens": 128 }
}

YOUR TOOLS

Pick where you build.

These examples use standard bearer authentication and OpenAI-compatible chat completions.

SETUP FORTerminal

export BINAH_API_KEY="paste-your-key-here"
export BINAH_BASE_URL="https://binah-gateway.binah-intelligenceworkersdev.workers.dev/v1"

curl "$BINAH_BASE_URL/chat/completions" \
  -H "Authorization: Bearer $BINAH_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "<callable-model-id>",
    "messages": [{"role": "user", "content": "Summarize this risk in three bullets."}],
    "stream": false
  }'

CONTROLLED-LAUNCH CONTRACT

Know what is—and is not—enabled.

Use exact model IDs. They make routing, price, and availability explicit.
binah/auto is intentionally disabled. No automatic-routing claim is made before qualification evidence exists.
Checking verified providers. The live serving catalogue is still loading.
No public checkout, contractual SLA, or GA claim. Access and support remain operator-managed.