FAQ

No. UltraCoder works out of the box with the built-in UltraSafe AIprovider — just download and open the app. You'll need to create an account and add credits to use the managed provider beyond the free tier. To use your own Anthropic, OpenAI, or other API key, go to Settings → Providers.

Yes — it's built on Microsoft Code-OSS 1.108, the open-source VS Code engine (MIT license). This means most VS Code extensions, themes, and keybindings work in UltraCoder, but it is a separate app and runs independently.

UltraCoder uses the VS Code extension API, so many extensions are compatible. You can install extensions via the Extensions panel (⌘⇧X). Some marketplace extensions may require the official VS Code marketplace; UltraCoder uses the Open VSX registry by default.

Code and file content are sent to the AI provider you've selected to generate responses. If you use the UltraSafe managed provider, requests go through UltraCoder's proxy servers. If you use your own API key (Anthropic, OpenAI, etc.), requests go directly from the IDE to that provider — not through UltraCoder. Local model providers (Ollama, LM Studio) never send data off your machine.

Approval is a safety mechanism. UltraCoder has access to your file system and terminal — requiring explicit approval prevents accidental overwrites or unexpected commands. If you trust UltraCoder for a task, you can enable auto-approval for specific tool types in Settings.

Ghost Suggestionsare real-time, line-level completions that appear as you type — like Copilot. They don't have memory of the broader task. The AI chat handles multi-step tasks: it reads your files, writes edits, runs commands, and works through complex requests with full context.

Open the Settings panel in the AI panel → Providers → select UltraSafe AI→ set as default. If you've run out of credits, add more from the Dashboard.

The IDE itself works offline, but AI features require an internet connection for cloud providers. To use AI offline, configure a local model via Ollama or LM Studio — both run entirely on your machine. See Models & Providers.

Use the Report Bug command in the Command Palette, or the report_bug tool available in any task. For account or billing issues, contact support from the dashboard.