Connect ChatGPT and Claude to your business processes

What you connect with ChatGPT and Claude

• Summarise and classify incoming email and tickets
• Extract data from invoices and documents for your accounting
• Draft replies to customer questions, set aside for review
• Enrich and prioritise leads and requests
• AI agents that work in your systems with permissions per action

How the integration works

A model only gets the data the task needs and must answer in a fixed format. That answer is checked before anything is done with it. Models run through a gateway, so you can switch provider and every call is logged and capped. For sensitive data, a local model can take over the task.

Facts about the ChatGPT and Claude API

Checked on 2 November 2026 against the official documentation. Limits, prices and versions change; the source is listed per row.

Fact	What the documentation says	Source
Authentication	API key as Bearer token: per project at OpenAI, per workspace at Anthropic with a mandatory version header.	OpenAI: Authentication
Rate limits	OpenAI: per project and per model, in calls and tokens per minute, with ascending usage tiers. Anthropic: per organisation with the tiers Start, Build and Scale. Both return 429 with a wait time.	OpenAI: Rate limits
Anthropic limits	The Start tier allows for example 4,000 calls per minute and a monthly spend cap of $500; higher tiers on request.	Anthropic: Rate limits
Data retention	OpenAI keeps abuse monitoring logs for up to 30 days by default; Zero Data Retention on request. Anthropic deletes within 30 days by default; Zero Data Retention via sales, not for the Fable and Mythos models without authorisation.	Anthropic: API and data retention
Training	Neither uses API data to train models unless you opt in yourself.	OpenAI: Your data
EU processing	OpenAI: EU data residency (storage and processing) per project through eu.api.openai.com, requested via sales. Anthropic: the first-party API offers only us or global, no EU option; through Amazon Bedrock or Google Cloud EU regions are available.	Anthropic: Data residency
n8n	Built-in OpenAI node (Responses API) and Anthropic node, plus chat model sub-nodes for AI agents.	n8n documentation

What you need on your side

• Your own OpenAI or Anthropic account with an API key in your company's name; the costs run through you
• A decision on which data may go to an external model and which stays local
• Examples of real tasks (emails, documents, tickets) to test the model on
• A monthly budget as a cap on the calls

Where it often goes wrong, and how I prevent it

• Convincing but wrong: a language model can make things up. That is why fixed logic checks the outcome, and whatever does not add up goes to a person.
• Costs: every call costs money. The integration works with a ceiling and alerts you when it comes into view.
• Privacy: not all data belongs with an external AI provider. Sensitive tasks go to a local model.
• Dependency: without a gateway you are tied to one provider. With a gateway, switching is a configuration choice.

In practice

In my own trading system BH28, ChatGPT and Claude help with review, analysis and reporting, but they never get the power to place orders: a model answer only counts after fixed validation. I build the same boundary for client processes with financial consequences.

What it costs

I build a process with ChatGPT or Claude as one scoped process for a fixed price, from €750. Maintenance and monitoring are a monthly fee: n8n management from €250 per month, server management €145 per server. You pay the model costs per token directly to the provider; in the free assessment we estimate them up front. All prices and what is included →

What I do not do

• I do not build my own models and do not train models on your data.
• I do not let a model carry out irreversible actions without a fixed check or approval.
• I do not promise EU processing where the provider does not offer it; then we choose a local model or a cloud platform with an EU region.

Frequently asked questions about connecting ChatGPT and Claude

Does my data go to OpenAI or Anthropic?

Only the data you send along for a task, and only if you choose to. For sensitive data, a model on your own server can do the task.

Can AI make decisions in my systems by itself?

Only within the limits you set. Write actions with consequences go through fixed checks or an approval; every action is logged.

What if a model is not available?

Then the workflow falls back to a route without a model, or sets the task aside for a person. Nothing gets lost.

Does my data stay in the EU with OpenAI or Anthropic?

With OpenAI it can: a project with EU data residency for storage and processing, requested via sales. Anthropic's own API offers no EU option; then you choose Claude through Amazon Bedrock or Google Cloud with an EU region, or a local model.

Does OpenAI or Anthropic train on my data?

Not through the API, unless you opt in yourself. Both keep data for at most 30 days by default for abuse monitoring; with Zero Data Retention that goes away too, on request and with conditions.