Skip to content

Repository files navigation

Tempest — parallel AI agent sessions

The token-efficient, open-source alternative to Conductor.build.

Run Claude Code, Aider, OpenCode, Copilot CLI, and more in parallel. Each isolated in its own git worktree and branch. Zero merge conflicts, live status, built-in diff and PR.

Version Downloads macOS Windows Linux Tauri License CI

Tempest — parallel AI agent sessions

Why Tempest uses far fewer tokens

Run five agents in parallel and each one reads your entire codebase from scratch — the same files, the same context, five times over. You pay for every token, every time.

Token Intelligence is a local code-knowledge graph that lives on your machine and is shared across every parallel agent session. When an agent needs to understand your codebase, it pulls from the shared graph instead of scanning files on its own. The work is done once. Every session benefits.

  • Up to 64% less context token consumption
  • Up to 58% fewer tool calls

No other parallel-agent tool does this.

One window. Every agent. No collisions.

Claude Code, Aider, OpenCode, Copilot CLI, Cline, Goose — all running in parallel, each in its own isolated git worktree and branch. Agents never touch each other's files. No merge conflicts mid-run. No stashing. No detective work about who changed what.

A rogue agent run never touches your main branch or anyone else's work. Blast radius: zero.

  • Live status across every session — know the moment each agent finishes, without babysitting.
  • Full history per session — close a tab, reopen it, the agent picks up exactly where it left off.
  • Built-in diff and PR — review each agent's changes, then stage, commit, push, and open a PR without leaving Tempest.

Process-isolated. Fully autonomous. Out of the box.

Every agent session in Tempest runs inside Hephaestus — a first-party process isolation layer built into the app. No configuration required.

Platform Isolation
Windows Job Objects — entire process tree confined and killed atomically on session close
macOS Seatbelt (SBPL) — deny-default sandbox via sandbox-exec
Linux bubblewrap — --unshare-pid --die-with-parent --unshare-net user namespaces

Agents also run fully permissionless by default. Tempest passes each agent's skip-permissions flag at spawn — no mid-run confirmation dialogs, no interruptions. Claude Code gets --dangerously-skip-permissions, Gemini CLI gets --yolo, Codex CLI gets --dangerously-bypass-approvals-and-sandbox.

Both behaviours are toggles in Settings → Security. You stay in control.

Tempest is built using Tempest — every feature in this repo was shipped by parallel agents running inside the app.

What's next

Database Branches — per-project Docker-based DB isolation is shipping now. Each project gets its own branched database container so agents can migrate, break, and restore data without touching anything else.

See ROADMAP.md for the full picture. Star this repo — we announce here first.

Build from source

Pre-built binaries are available for Windows, macOS, and Linux.

# Prerequisites: Node.js 18+, Rust 1.77+
# Windows also requires WebView2 Runtime:
# https://developer.microsoft.com/en-us/microsoft-edge/webview2/
git clone https://github.com/tempestai-dev/tempest
cd tempest
npm install
npm run dev        # development with hot reload
npm run build      # production build -> dist-installers/

Community

X (Twitter) — @usetempest

GitHub

Instagram

LinkedIn

About

Run Claude Code, Codex, Gemini and any other CLI Agent with 64% fewer tokens and 54% fewer tool calls

Topics

Resources

Code of conduct

Contributing

Stars

Watchers

Forks

Releases

Contributors

Languages