Integrating Claude and ChatGPT APIs Into Legacy ColdFusion Applications

ColdFusion is exceptionally well-suited to calling AI APIs, because both Claude (Anthropic) and ChatGPT (OpenAI) are plain HTTP/REST + JSON services — and cfhttp, serializeJSON(), and deserializeJSON() are exactly the tools CFML has always had. You don't rewrite your legacy app; AI integration is additive. You POST a JSON request to the provider's endpoint with your API key in a header, parse the JSON response, and use the result. For Claude, that's POST https://api.anthropic.com/v1/messages with an x-api-key header and the anthropic-version: 2023-06-01 header. For ChatGPT, it's POST https://api.openai.com/v1/chat/completions with an Authorization: Bearer header. The one ColdFusion-specific gotcha that trips everyone up: LLMs are slow (they generate token by token), so you must set cfhttp timeout high and set the page requesttimeout higher still, or ColdFusion kills the thread before the model responds. Newer ColdFusion (the 2025 release) even ships a native ChatModel() abstraction, but the cfhttp approach works on any CF version back to legacy installs. This guide shows both providers, both syntaxes, and the production concerns that matter.
Read More