AI Gateway quickstart

Send one chat completion to deepseek/deepseek-v4.1-flash and read the credits it charged.

Before you start

Copy your key from the dashboard under Integrations → API Keys and store it as PREDEV_API_KEY. Keys look like pdk_…. A free workspace can make 5 credits of gateway calls; after that, calls need a subscription or credits. See API key and pricing.

Send a request

curl
Node.js
Python
Vercel AI SDK
Anthropic SDK

curl --fail-with-body https://api.pre.dev/v1/chat/completions \
  -H "Authorization: Bearer $PREDEV_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "deepseek/deepseek-v4.1-flash",
    "messages": [{ "role": "user", "content": "Say hello in five words." }]
  }'

Install the stock SDK with npm install openai, then set baseURL:

import OpenAI from 'openai';

const client = new OpenAI({
  baseURL: 'https://api.pre.dev/v1',
  apiKey: process.env.PREDEV_API_KEY!,
});

const completion = await client.chat.completions.create({
  model: 'deepseek/deepseek-v4.1-flash',
  messages: [{ role: 'user', content: 'Say hello in five words.' }],
});
console.log(completion.choices[0].message.content);

Install the stock SDK with pip install openai, then set base_url:

import os
from openai import OpenAI

client = OpenAI(
    base_url="https://api.pre.dev/v1",
    api_key=os.environ["PREDEV_API_KEY"],
)

completion = client.chat.completions.create(
    model="deepseek/deepseek-v4.1-flash",
    messages=[{"role": "user", "content": "Say hello in five words."}],
)
print(completion.choices[0].message.content)

Install ai and @ai-sdk/openai, then create a provider with the gateway URL:

import { createOpenAI } from '@ai-sdk/openai';
import { generateText } from 'ai';

const predev = createOpenAI({
  baseURL: 'https://api.pre.dev/v1',
  apiKey: process.env.PREDEV_API_KEY!,
});

const { text } = await generateText({
  model: predev('deepseek/deepseek-v4.1-flash'),
  prompt: 'Say hello in five words.',
});
console.log(text);

The gateway serves the Anthropic Messages shape at /v1/messages. The Anthropic SDK appends /v1/messages itself, so its base URL is https://api.pre.dev without /v1. Use your pre.dev key as the API key:

import Anthropic from '@anthropic-ai/sdk';

const client = new Anthropic({
  baseURL: 'https://api.pre.dev',
  apiKey: process.env.PREDEV_API_KEY!,
});

const message = await client.messages.create({
  model: 'anthropic/claude-sonnet-5',
  max_tokens: 256,
  messages: [{ role: 'user', content: 'Say hello in five words.' }],
});
console.log(message.content);

Model ids stay in catalog form (anthropic/claude-sonnet-5), not Anthropic’s native names. See Messages.

Read the response

The body is a standard chat completion response. Three headers come from the gateway:

Header	Meaning
`x-predev-request-id`	Gateway request id, `pdr_…`; quote it in support requests
`x-predev-credits-charged`	Credits charged for this call
`x-predev-credits-remaining`	Workspace balance after the call; omitted on unlimited plans

Add -i to the curl command to print them. In the SDKs, use the raw-response helpers (client.chat.completions.with_raw_response in Python, .withResponse() in Node) when you need headers.

Swap the model

Reasoning models spend your max_tokens thinking. The gateway enables reasoning by default on models that support it (Claude, DeepSeek, Gemini and others). With a small max_tokens the reply can come back as content: null with finish_reason: "length" because the whole budget went to reasoning tokens — which are still billed. For chat, summarise or classify features either pass reasoning: { "enabled": false } (or { "effort": "low" }) or give max_tokens at least 1024.

Change model to any id from GET /v1/models, for example anthropic/claude-sonnet-5, google/gemini-3.8-flash, or openai/gpt-5.6-luna. An invalid id returns the provider’s error with its original status and message. Continue with streaming, tools and structured output, or routing and fallbacks.

Get started

Guides

API reference

AI Gateway quickstart

Before you start

Send a request

Read the response

Swap the model

​Before you start

​Send a request

​Read the response

​Swap the model

Before you start

Send a request

Read the response

Swap the model