Frequently asked

The app is free and the code is on GitHub. What you pay for is the model usage at the provider you chose, or nothing at all with a local model through Ollama or LM Studio.

Websites answer you. Rafiq works: it reads and writes your files, runs commands, browses, and carries out whole tasks on your folder, with every sensitive step going through your approval.

Twenty-one providers: Anthropic, OpenAI, Google Gemini, DeepSeek, Mistral, xAI, OpenRouter, GitHub Copilot, Azure OpenAI, AWS Bedrock, Google Vertex AI, Groq, Cerebras, Fireworks, Together, Qwen, Kimi and GLM, plus Ollama and LM Studio locally, and any OpenAI-compatible server.

With a local model (Ollama or LM Studio), yes: files, commands and designs are all local. The web tools naturally need a connection.

Only what the model asks for to do your job, and straight to the provider whose key you added. There is no Rafiq server in between, no analytics, no tracking.

Only if you want to use GitHub Copilot models. You sign in with a GitHub account that has an active Copilot plan, without pasting any key.

Because the installer isn't signed with a code-signing certificate yet. Click More info, then Run anyway, and compare the file's hash with the SHA256SUMS.txt attached to the release if you want to be sure.

Windows 10 and 11 only for now. The architecture is ready for other systems, but there is no published build for them yet.

From the providers' own rate tables. A model with no known price is counted in tokens only; it never invents a number.

Open an issue on GitHub and attach the diagnostics report from Settings → Usage: a JSON file with the app version, your settings and provider names, with no key and no conversation content.