How to Get an API Key for AI

In DEVONthink 4, you can use an external AI model, e.g., to ask questions about your documents or generate images. To integrate an AI model such as ChatGPT or Claude, you need an API key. But what is this and where can you get one?

An API key — API is an abbreviation for application programming interface — is a special key, unique to you and issued by the AI provider. It allows you to use their services in DEVONthink. In Settings > AI > Chat you will see a field marked API Key where you can add your key.

API keys are often separate from chat accounts, so if you are chatting with Claude, this doesn’t mean you have an API key. But you can just ask the AI model, “How do I get an API key for your services?” It should get you pointed in the correct direction. Alternatively, visit the website of the AI provider, where you should find a separate section for API keys.

Note, the use of a commercial API is usually not free, but it’s likely you can set up a pay-as-you-go plan, so you’re only billed when actually using the service. We discuss this in the Getting Started > AI Explained section of the built-in Help and user handbook.

Also, be aware you may need more than one key, depending on the service you use. In the AI > Image Generation settings, you’ll need to enter an API key to use. However, OpenAI’s image generators and transcription model use the same key as your chat key, so you can simply use this one. For other image engines, an API key from Replicate.com gives you access to several excellent options.

Another option is to use local models via engines like Ollama or LM Studio. We will write a separate blog post about this.