Skip to content
OrqLabs
Türkçe'ye geç: ENTR

Your own model keys

OpenAI API key (BYOK): GPT models, image generation and knowledge base

Connect your OpenAI key and three things run on your own OpenAI account: agents and AI nodes set to a GPT model, the **Generate image** node, and embeddings for knowledge base documents.

Start freeStep-by-step setup guide
Connection type
API key
Scope
Per organization: all projects share one key
Capabilities
your own key
Setup time
10 min

What it does in OrqLabs

Using your OpenAI API key unlocks GPT models alongside Claude in OrqLabs: GPT-5, GPT-5 mini, GPT-5 nano, GPT-4.1, GPT-4.1 mini and GPT-4o are ready in the catalog (o3 and o4-mini are off by default). Pick a GPT model in an agent or AI node and the call goes to OpenAI.

The key isn't only for text models. The Generate image node runs on OpenAI's image model (Claude doesn't generate images), and knowledge base documents are indexed with text-embedding-3-small, which also powers knowledge base search. So even if your agents only use Claude, the knowledge base and image generation need an OpenAI key, either the platform's or yours.

Billing follows the same rule as Anthropic BYOK: calls are billed to your OpenAI account, OrqLabs adds no markup and the spend doesn't count toward your plan's monthly AI budget, while usage is still metered and per-run budgets apply. Images are priced per image rather than per token.

What you can automate

  • Post images

    In the daily social content template, the image prompts written by the Content Creator go to the Generate image node, which produces 1024x1024, 1024x1536 or 1536x1024 images and sends them for your approval.

  • A chatbot backed by your knowledge base

    Upload FAQs, price lists and policy documents to a knowledge base; they're indexed with OpenAI embeddings, and the AI Chatbot finds the relevant passages while answering on WhatsApp or Telegram.

  • Economy models for high volume

    Choose economy-tier models such as GPT-5 nano or GPT-4.1 mini in Classify, Extract structured data or Generate text nodes to run classification and short-text jobs at low cost.

  • Long documents

    For nodes that process very long text, pick GPT-4.1 with its context window of roughly one million tokens.

Nodes that use this integration

Every agent and AI node can run on this key.

In the workflow editor these nodes run with your OpenAI API key (BYOK) account.

  • Generate text

    Has the AI write text from your instructions in a single step, e.g. a short reply draft for every new comment.

  • Extract structured data

    Turns free text into tidy fields such as name, phone and budget, e.g. turning a WhatsApp booking request into a record.

  • Classify

    Sorts each text into one of your categories and sends it down that path, e.g. splitting incoming messages into sales, support and spam.

  • Translate

    Translates the text fields you choose into another language, keeping your brand tone, e.g. an English version of a post for LinkedIn.

Setup at a glance

  1. Add a balance at platform.openai.com and, if you like, create a separate OrqLabs project with its own budget and model access.

  2. Create and copy a service account key under API keys → Create new secret key.

  3. In OrqLabs, open Integrations → OpenAI API key (BYOK) → Connect.

  4. Enter the key in API key; fill in Organization ID only for legacy keys, then click Connect.

  5. Check it with Test connection; from then on, GPT calls, image generation and new document indexing use this key.

Read the full setup guide →

Requirements

  • A plan that includes BYOK (Starter and up); it's disabled on the FREE plan.
  • An OpenAI Platform account with a payment method and a positive balance.
  • A project key (sk-proj-…); restricted keys need the Model capabilities: Write permission. Most setups don't need an Organization ID.
  • OpenAI may ask some accounts to verify their organization before using image models.

What you enter when connecting

  • API key *
  • Organization ID (optional)

Good to know

  • BYOK doesn't widen plan tiers; flagship models such as GPT-5 are available only if your plan includes that tier.
  • The embedding model is set platform-wide; if it changes, knowledge base documents have to be re-indexed.
  • When your balance runs out, OpenAI rejects calls and retrying won't help; top up or turn on auto recharge in OpenAI.

Frequently asked questions

I only use Claude. Do I still need an OpenAI key?

If you use the knowledge base or image generation, yes: either the platform's OpenAI key or your own BYOK key. If you only generate text with Claude, you don't.

Do I have to re-index existing documents after connecting?

No. Indexing and searches after you connect use your key, and existing documents keep working as long as the embedding model stays the same.

How much does image generation cost?

OpenAI prices images per image; check OpenAI's price list and your usage dashboard for current rates.

Which type of OpenAI key should I use?

A project key owned by a service account is recommended; keys owned by a person stop working when that person is removed from the OpenAI organization.

Get started with OrqLabs

Build your first workflow today

Start on the free plan: connect your accounts, pick a template and stay in control with approval steps. No credit card required.

Start freeExplore the product