Migrating from OpenRouter

OpenRouter gave you multi-provider access behind one key. Behest gives you that plus per-end-user identity, session memory, BYOK at wholesale rates, and browser-safe tokens. The migration itself is a baseURL change.

Step 1: add your provider keys

OpenRouter pays providers for you from their wallet. Behest supports two modes:

1. BYOK — paste your OpenAI / Anthropic / Google keys in dashboard → Providers. You pay providers directly at their list prices, no markup.
2. Behest-hosted — use Behest's pool at wholesale (currently OpenAI + Anthropic), billed on your Behest plan.

You can mix: BYOK OpenAI, Behest-hosted Anthropic, etc. Behest routes per model.

Step 2: swap the client

Heads-up — Behest is not a one-header swap. OpenRouter accepts your long-lived sk-or-v1-... as the Bearer on /api/v1/chat/completions. Behest does not: Kong requires a minted JWT (short-lived, per end-user). Your behest_sk_live_... only authenticates the mint endpoint. Replacing the OpenRouter key with BEHEST_KEY directly in the OpenAI SDK will 401. Either use the v1.5 Behest SDK (which mints for you), or mint manually and pass the JWT.

Node / TypeScript

Before (OpenRouter):

const client = new OpenAI({
  apiKey: process.env.OPENROUTER_API_KEY,
  baseURL: "https://openrouter.ai/api/v1",
});

After — recommended: v1.5 Behest SDK

import { Behest } from "@behest/client-ts";
const behest = new Behest({
  key: process.env.BEHEST_KEY, // behest_sk_live_...
  baseUrl: `https://${BEHEST_SLUG}.behest.app`,
});
// SDK calls /v1/auth/mint per request, injects the JWT, streams back.
await behest.chat.completions.create({ messages, user_id: userId });

After — raw OpenAI SDK: mint the JWT on the server, hand it to the client as apiKey.

import OpenAI from "openai";
 
const mint = await fetch(`https://${BEHEST_SLUG}.behest.app/v1/auth/mint`, {
  method: "POST",
  headers: {
    "Content-Type": "application/json",
    Authorization: `Bearer ${process.env.BEHEST_KEY}`,
  },
  body: JSON.stringify({ user_id: userId, session_id: crypto.randomUUID() }),
}).then((r) => r.json());
 
const client = new OpenAI({
  apiKey: mint.access_token, // JWT — NOT BEHEST_KEY
  baseURL: `https://${BEHEST_SLUG}.behest.app/v1`,
  defaultHeaders: { "X-Session-Id": mint.session_id },
});

Python

# Before
client = OpenAI(api_key=OPENROUTER_KEY, base_url="https://openrouter.ai/api/v1")
 
# After — using the v1.5 Behest SDK (mints for you):
from behest import Behest
behest = Behest()  # reads BEHEST_KEY + BEHEST_BASE_URL from env
behest.chat.completions.create(messages=..., model="gpt-4o-mini", user_id=user_id)
 
# After — using raw OpenAI SDK: mint first, pass the JWT.
import uuid, requests
from openai import OpenAI
mint = requests.post(
    f"https://{SLUG}.behest.app/v1/auth/mint",
    headers={"Authorization": f"Bearer {BEHEST_KEY}"},
    json={"user_id": user_id, "session_id": str(uuid.uuid4())},
).json()
client = OpenAI(
    api_key=mint["access_token"],              # JWT — NOT BEHEST_KEY
    base_url=f"https://{SLUG}.behest.app/v1",
    default_headers={"X-Session-Id": mint["session_id"]},
)

Model name mapping

OpenRouter prefixes provider names (openai/gpt-4o, anthropic/claude-3.5-sonnet). Behest accepts either:

If you have OpenRouter as a BYOK provider in Behest, all openrouter/* models keep their original names — pass through unchanged.

You can also set a project default model and drop the model field entirely; Behest fills it in.

Feature parity

Routing and fallback

OpenRouter lets you specify a provider order per request:

{ "provider": { "order": ["OpenAI", "Azure"] } }

Behest's equivalent is project-level fallback chains in the dashboard:

• Primary: gpt-4o via your OpenAI BYOK
• Fallback 1: claude-3-5-sonnet via Anthropic BYOK
• Fallback 2: Behest-hosted gpt-4o

Configure once; every request inherits the chain. Or pass model per request to pin a specific route.

Headers and extras

OpenRouter-specific headers (HTTP-Referer, X-Title) are ignored by Behest — not needed for attribution. Replace with:

• X-Session-Id for ephemeral memory (set automatically by the SDK after behest.auth.mint(), or pass session_id per chat.completions.create call)
• X-Thread-Id for persistent conversations
• A per-user Behest JWT (via behest.auth.mint({ user_id })) for rate limits

Common upgrade reasons

• "My OpenRouter key leaked in a browser bundle" → Behest JWTs expire in 15 min and are scoped to one user. Not a catastrophic leak.
• "One abusive user exhausts my rate limit" → Behest per-user RPM prevents this by construction.
• "I need conversation history but don't want to store messages myself" → Use Behest threads.
• "I want to show users their usage" → GET /v1/billing/usage filtered by uid.

See also

• Authentication
• BYOK deep dive
• Providers
• Migrating from OpenAI