Models

Choosing a model

Pick the provider in the header of the Lucicoder panel, and the model in the searchable model picker. The picker shows every model the provider currently offers, with badges for tool calling and vision, and says where the list came from.

• Local, through Ollama — works fully offline; nothing leaves your computer.
• Claude Code or Codex — uses your Claude or ChatGPT subscription through the vendor's official command-line tool.
• A cloud provider with your key — Anthropic, OpenAI, Google Gemini, DeepSeek, Mistral, Groq, xAI, OpenRouter, DashScope, or any OpenAI-compatible endpoint.

Local models with Ollama

1. Install Ollama.
2. Pull a coding model:

ollama pull qwen2.5-coder:7b

3. In Luciai Studio, pick Ollama as the provider in the Lucicoder panel, then the model. Lucicoder finds the models you pulled on its own.

qwen2.5-coder is a good start; llama3.1 and the other Ollama models work too. With a local model, your prompts and your code stay on your machine.

Small models and tool calls

Lucicoder uses each provider's native function calling — for Ollama, its tools API — and falls back automatically to a universal XML grammar, so even small local models work reliably. You can choose yourself with /native:

• /native auto — detect what the model supports.
• /native on — always use native function calling.
• /native off — always use the XML grammar.

Ghost-text completions

Inline suggestions as you type come from a local fill-in-the-middle model through Ollama's FIM protocol — qwen2.5-coder, codellama:code, deepseek-coder, starcoder2 or codegemma — or from any other provider by instruction. Tab accepts a suggestion. Switch them on or off in the status bar; choose the model, and opt out per language, in Settings → AI.

Your subscription: Claude Code and Codex

If you have a Claude or ChatGPT subscription, Lucicoder can work through it: switch the engine to Claude Code or Codex, and Lucicoder drives the official CLI you are logged into.

1. Switch Lucicoder's engine to Claude Code or Codex.
2. If the CLI is not installed or not logged in yet, use the install and login buttons. They run the exact commands in the terminal, so you see what happens.
3. Work as usual. Lucicoder sends your requests through the CLI.

Cloud providers with your own keys

Choose Anthropic, OpenAI, Google Gemini, DeepSeek, Mistral, Groq, xAI, OpenRouter or DashScope — or any OpenAI-compatible endpoint, such as a model server on your own network. Add your API key in the provider settings, which open from the model picker.

Already have the vendor's usual environment variable set, such as ANTHROPIC_API_KEY, OPENAI_API_KEY or GEMINI_API_KEY? Luciai Studio picks it up automatically.

Model lists are fetched live from each provider, so a new model appears the day it ships. The lists are cached, and when a provider cannot be reached, a built-in list is used and labelled as such.

Cloud usage is billed by the provider, directly to you. We are not in between.

Where your keys are kept, and what is sent

API keys and tokens are kept in your operating system's credential store — on Windows, the Credential Manager. They are not written into your project, and they are never sent to us.

• With a local model, prompts and code stay on your machine.
• With a cloud provider, they go to that provider only, under its terms.
• With Claude Code or Codex, they go through that vendor's CLI to that vendor.

We never receive your code. Privacy and data lists everything that does leave your machine.