Built for developers who run agents at scale. Locally. In parallel.

MISSION

Tempest exists to make running multiple AI coding agents practical — not in theory, but in daily development work. The core problems are isolation and cost. Without isolation, agents overwrite each other's changes. Without shared context, every agent pays the full token cost of reading your codebase from scratch.

Tempest solves both. Each agent session runs on its own git worktree — a separate working directory linked to your repo. Token Intelligence builds a local knowledge graph of your codebase once and shares it across every parallel session, cutting context consumption by up to 64%.

HOW IT IS BUILT

Tempest is a desktop app built with Tauri 2.x — a Rust backend with a React frontend running in a native WebView. It ships native binaries for Windows, macOS, and Linux. No Electron, no bundled Chromium. The app is small, fast to start, and runs entirely on your machine.

The PTY layer handles real terminal sessions for each agent. Git worktrees handle isolation. Token Intelligence (powered by Atlas) handles the shared code-knowledge graph. Nothing leaves your machine — not your code, not your context, not your agent conversations.

OPEN SOURCE

Tempest is free and open source under the Apache 2.0 license. Use it, fork it, build on it, ship a modified version commercially — no restrictions. The source is on GitHub. Issues, pull requests, and feedback are welcome.

STATUS

Tempest is in active early development. Windows binaries are available now. macOS and Linux binaries are on the roadmap. Token Intelligence shipped in v0.1.2. Every release is documented in the release notes.