Skip to main content
Coreply connects to any OpenAI-compatible inference API. That means you can use hosted providers like Google AI Studio, Groq, or OpenAI, as well as any self-hosted endpoint that follows the OpenAI API format.

​
What you need

To connect Coreply to a provider, you need three things:
  • API URL — the base endpoint for the provider’s inference API
  • API Key — your authentication credential from the provider
  • Model Name — the identifier of the model you want to use
Each provider page below shows you exactly where to find these values.

​
Choosing a model

Model size has a meaningful impact on output quality with Coreply. Smaller models under 7 billion parameters often struggle to follow Coreply’s output format, which can result in poor or unusable suggestions. Models at 7B parameters and above generally work well. Reasoning models — such as GPT-5, Gemini 2.5, and Claude 3.7 and above — are technically compatible, but they introduce extra latency before each suggestion appears. For a real-time texting experience, non-reasoning models are a better fit. Some providers like OpenRouter disable reasoning by default, which can make these models usable.

​
Supported model families

The following model families have been tested and confirmed to work with Coreply’s prompt format:
ModelSupported
Claude 3 & 4 Family✅
Llama 3 & 4 Family✅
Gemma 2 & 3 Family✅
OpenAI GPT Family✅
Google Gemini Family✅
Many other models should also work. You’re welcome to experiment with models not listed here.

​
Provider setup guides

Google AI Studio

Use Gemini models via the free AI Studio API.

Groq

Fast inference on open-source models with a free tier.

OpenRouter

Access hundreds of models from a single API key.

OpenAI

Use GPT-4.1 and GPT-4o models directly from OpenAI.

Mistral

Use Mistral models, including FIM mode with Codestral.

Custom provider

Any OpenAI-compatible endpoint works. Find the API URL, API Key, and Model Name in your provider’s docs.