Using Gab AI API with Cursor IDE

Connect Cursor IDE to Gab AI's OpenAI-compatible API and use any of our models for chat, composer, agent mode, and inline edits.

Prerequisites

Configuration

Cursor Settings → Models → Add Custom Model. Choose OpenAI Compatible for most models (base URL https://gab.ai/v1). For Claude models, choose Anthropic protocol and set the base URL to https://gab.ai (no /v1). Paste your Gab AI API key. Set the Agent / Composer model to auto so Auto Router classifies the kickoff and sticks that model for the tool loop. Pin a concrete slug for Tab and inline complete — those must not use Auto. Copy context_window and max_output_tokens from the catalog into Cursor's context length / max output fields, and enable tool calling. GET https://gab.ai/v1/agent-setup?client=cursor returns a generated model list.

Selecting a Model

After verification, you can use any text-capable model from your Gab AI account. Set the model name to the model's slug or an alias from GET /v1/models (e.g. arya, gpt-5-5, gpt-5.5, claude-sonnet-4-5, anthropic/claude-sonnet-4-5). Display names like "Claude Opus 4.8" will not match.

List All Models

You can retrieve the full list of available models by calling GET /v1/models.

What Works in Cursor

Limitations

Tab Completion

  1. Open Cursor Settings → Models → Add Custom Model — Open Cursor Settings → Models → Add Custom Model
  2. Choose OpenAI Compatible (or Anthropic for Claude slugs) — Choose OpenAI Compatible (or Anthropic for Claude slugs)
  3. Set the OpenAI base URL to https://gab.ai/v1, or https://gab.ai for Anthropic protocol — Set the OpenAI base URL to https://gab.ai/v1, or https://gab.ai for Anthropic protocol
  4. Paste your Gab AI API key and a model slug from GET /v1/models — Paste your Gab AI API key and a model slug from GET /v1/models
  5. Fill context length and max output from the catalog; enable tool calling — Fill context length and max output from the catalog; enable tool calling

List All Models

You can retrieve the full list of available models by calling GET /v1/models.