Guides

AI models and access

Elmony can build with an AI subscription already available on your computer, an API provider, Elmony AI credits, or a supported local model. You only need to configure one working access path to get started.

Choose an access path

Recommended if you already subscribe

Subscription Agents

Use Claude Code, Codex with your ChatGPT/OpenAI account, or Gemini CLI. These are local command-line agents that Elmony connects to for Builder Chat and Elmony Design.

Open Settings and select Subscription Agents.
Install any missing CLI or adapter from its card, then complete the provider's sign-in flow.
Click Refresh, choose a model or reasoning option, and select Use as default.

This setup is machine-wide. The subscription provider controls account eligibility, usage limits, available models, and billing.

Alternative

Model Providers

Connect a provider with your own credentials, or select Elmony AI to spend credits from your signed-in Elmony account without adding a provider API key.

Open Settings and select Model Providers.
Open a provider card and save the requested API key or provider credentials.
Return to Builder Chat and choose one of its models from the model picker.

API usage is billed by the selected provider unless the model picker explicitly says it uses Elmony AI credits.

Where AI settings live

AIThinking, context, and maximum chat-turn defaults.

Subscription Agents Installation, sign-in status, model/reasoning choice, and the default CLI agent.

Model Providers API credentials, provider-specific configuration, and custom models.

Tools (MCP) Model Context Protocol servers and per-tool permission rules.

Provider families

The exact model catalog changes as providers release and retire models. Depending on your Elmony version and account, Model Providers can include direct services such as OpenAI, Anthropic, and Google; gateways or cloud platforms such as OpenRouter, Google Vertex AI, Azure OpenAI, AWS Bedrock, and xAI; and Elmony AI credits. Use the model picker as the source of truth for models currently available to you.

Add a custom provider Use local models

Models, modes, and tools

Choosing a model controls which AI handles the request. Choosing a chat mode controls how Elmony uses it. Build New is the recommended full workflow for both API models and subscription agents, while the other modes are useful for direct iteration, questions, classic edits, or MCP tools.

See Modes, tools, and context for Build New, Build, Simple, Ask, Build with MCP, Smart Context, Web Access, and Turbo Edits.

Pricing and limits change

Do not rely on a fixed daily request count. Free tiers, rate limits, model access, and prices vary by provider, account, region, and plan. Check the usage and billing dashboard for the provider or subscription agent you selected before starting a large build.

Choosing efficiently

Start with the default model or reasoning option for your access path. Move to a stronger option for architecture, difficult debugging, or broad refactors; use a faster option for focused edits and routine follow-ups. The model usage guide explains how modes and context affect usage.