FAQ

What is nala?

nala is an agentic IDE: a private, local-first workspace where your AI coding agents run side by side. It combines a durable terminal multiplexer, a local orchestration layer, on-device voice, and MCP integration in one product. The overview covers the three surfaces, and how nala works explains the daemon underneath them.

Which coding agents does nala support?

Claude Code, Codex CLI, Devin, Cursor, Grok Build, Gemini CLI, Kimi Code, OpenCode, and any custom CLI agent you define. See the providers section for connection guides and the capability matrix.

Does nala see my API keys?

No. Keys you supply travel from your machine straight to that provider: never stored on our servers, never proxied, never logged, never included in a support bundle. nala does not sign in to providers on your behalf, and a provider subscription cannot be routed through us. Credentials documents exactly where secrets live.

Do I need to pay for AI inference?

nala does not sell model inference. The app is free; any charges for a third-party coding tool or model account remain between you and that provider. An optional managed plan is planned for people who would rather not manage keys; the billing section describes it, and nothing there is purchasable today.

Is nala really free? What is the catch?

The desktop app, terminal agent, CLI, every agent integration, on-device voice, and updates are free, and stay free. That is the whole product, not a trial of it. There is no account requirement and no meter running.

Which platforms does nala run on?

A Windows x64 build is published today; macOS and Linux builds are not yet published. Platform support lists exactly what exists, and installation covers system requirements.

The OS warned me about the installer. Is that expected?

Yes. Preview builds are currently unsigned, so Windows SmartScreen or macOS Gatekeeper will warn. The response is per-install, never a global bypass, and every release publishes the SHA-256 to verify against. The unsigned preview policy walks through it.

Does nala work offline?

Core workspace and on-device voice features are local-first and do not require a Nala website account. Third-party agents and model providers need a network connection under their own terms. Privacy covers what leaves the machine and what does not.

Can my team collaborate in nala?

You and your agents work in the same shared space: inspect tasks, review artifacts, and steer work without rebuilding context. Company mode (experimental) spins up teams of agents with cost tracking and approval gates. See company mode.

How do updates work?

nala checks the selected release channel, downloads in the background, verifies the SHA-256 before offering install, and aborts on any mismatch. Channels and the manual nala update command are covered under updates and channels.

What does the website collect about me?

Only with your consent. If you allow analytics, the site records coarse page, campaign, browser, and download events under a random 30-day session; it never stores raw IP addresses, full user agents, code, prompts, or project names. If you decline, nothing is sent. The product itself has no telemetry.

What is on the roadmap?

The immediate roadmap is a trustworthy cross-platform release and update path. Hosted services are planned, not offered. Feature status tracks what ships today, and the changelog records what landed.

How do I uninstall?

See Uninstall for removing the app, and for where your data lives if you want that gone too.