tellnova is a desktop app that runs background coding agents. You describe a task, pick a repository, and tellnova works it inside an isolated git worktree while you keep coding. When the agent finishes, you review the diff and open a pull request.

Which AI models and providers does tellnova support?

tellnova runs on the pi coding agent and supports your existing subscriptions and API keys: Codex (ChatGPT), Cursor, GLM (Z.ai), Kimi, MiniMax, and OpenRouter. You can pick a provider and model per task or per reply.

Yes. The tellnova desktop app is free during its public alpha for macOS (Apple silicon and Intel), Windows, and Linux. Model usage runs on your own provider subscriptions, and optional Tellnova Cloud sandbox execution is a paid add-on.

How does tellnova keep my repository safe?

Every task runs in its own git worktree on its own branch, so agents never touch your working copy. Credentials stay on your machine — tokens are passed through environment variables, never written into worktrees, and provider keys are stored in the app's private store.

Can several coding agents run at the same time?

Yes. tellnova is built for parallel work: each attempt gets an isolated worktree and branch, multiple tasks run concurrently, and the sidebar tracks every live thread with its status, terminal, and browser preview.

Does tellnova work without the desktop app?

Yes. tellnova ships a self-hosted server mode that exposes the same interface over authenticated HTTP, so you can drive tasks from a browser or your own machine while agents run on a server or home box.

tellnova — background coding agents in isolated git worktrees

tellnova — background coding agents in isolated git worktrees tellnova is a free desktop app that runs background coding agents. Describe a task, pick a repository, and an agent works it inside an isolated git worktree — parallel attempts, your own model subscriptions (Codex, Cursor, GLM, Kimi, OpenRouter), and one-click pull requests when the diff is ready. Download tellnova for macOS (Apple silicon and Intel), Windows 10/11, and Linux from the GitHub releases page. How it works

Choose the work — select a repository and describe the outcome.

Choose where it runs — your machine with your credentials, or an isolated cloud sandbox.

Review what returns — follow the run, inspect the diff, open a pull request.