Termic - Run Claude Code, Codex in Parallel Git Worktrees
A free, open-source desktop app that runs Claude Code, Codex, and other AI coding agents in parallel git worktrees. Built with Rust/Tauri, no vendor SDK required.
TL;DR
TL;DR: Termic is a free, open-source desktop app that runs real Claude Code, Codex, and other AI coding-agent CLIs in parallel git worktrees — no vendor SDK, no separate billing, just the Pro/Max plan you already pay for.
Source and Accuracy Notes
⚠️ This section is MANDATORY. All links must be verified from actual source, not guessed.
- Project page: termic.dev
- Source repository: github.com/simion/termic
- License: AGPL-3.0 (verified via GitHub API)
- HN launch thread: news.ycombinator.com/item?id=49056316
- GitHub metadata (verified via API): 125 stars, default branch
main, pushed 2026-07-28
What Is Termic?
Termic is a free, open-source desktop application for macOS, Linux, and Windows that runs AI coding-agent CLIs side by side, each isolated in its own git worktree. It spawns the real claude, codex, agy (Antigravity), copilot, and grok binaries — not vendor SDKs — so inference rides on the Claude Pro or Max plan you already pay for.
The key insight: since June 2026, Anthropic’s Agent SDK routes usage through a separate credit pool. Termic sidesteps this entirely by calling the CLI directly, the same way you would in iTerm.
Built with Rust and Tauri 2, with a React frontend, xterm.js with WebGL rendering, and a CodeMirror 6 editor.
Install
macOS (recommended)
brew install --cask simion/termic/termic
This auto-taps simion/homebrew-termic, downloads the latest .dmg, and installs Termic.app into /Applications. Termic is signed with a Developer ID certificate and notarized by Apple. Requires macOS 12+.
Updates are automatic: a pill appears in the top-right toolbar when a new release lands. Or manually:
brew upgrade --cask termic
Linux
Download the AppImage from the Releases page, make it executable, and run:
chmod +x termic_*_amd64.AppImage
./termic_*_amd64.AppImage
Keep it in a writable location like ~/Applications/ for the in-app updater to work.
On Wayland, if fonts render thin, force X11: GDK_BACKEND=x11 termic.
Build from source
git clone https://github.com/simion/termic
cd termic
make setup # installs rust, node, and cargo dependencies
make install # builds and launches
For a full development loop: make dev (Vite HMR + Rust auto-rebuild).
Key Features
Parallel git worktrees. Each workspace is a git worktree under ~/termic/workspaces/<project>/<name>/. Spin up four agents against the same branch across tabs, or attach an agent to the repo root when you do not want a worktree.
Bring your own agent. Settings → Agents is an editable registry. Drop in aider, ollama, or any CLI that runs in a terminal. Claude, Codex, Antigravity, Copilot, Grok, and opencode ship as built-ins.
Per-workspace sandbox (macOS only). Optional filesystem and network cage via sandbox-exec plus an in-process HTTPS CONNECT proxy with a hostname allowlist. Configured per project and enforced from agent spawn time. No external daemon.
Broadcast and Brainstorm. Send one prompt to every agent in a workspace concurrently via Shift+Cmd+B. Useful for multi-agent code reviews or getting parallel second opinions on a complex bug.
Spotlight. Mirror a worktree’s changes — committed, uncommitted, and untracked — into your repo root in real time, so your editor and browser see the agent’s work live. Runs on a detached HEAD and never commits to your branch.
Reliable work-done detection. Per-CLI title classifier plus OSC 9;4, gated by byte-quiet and content-hash checks. This enables opt-in desktop notifications that only fire when an agent actually finishes a turn.
Termic vs. Conductor
The most direct competitor is Conductor.build, which is closed-source and paid. The key differences, sourced from Termic’s own comparison page:
| | Termic | Conductor | |---|---|---| | License | AGPL-3.0 (open source) | Proprietary | | Price | Free | Paid | | Runs Claude CLI directly | Yes | Yes (but routes through SDK) | | Separate SDK billing on top of Pro/Max | No | Yes | | Multi-repo workspaces | Yes | No | | Attach to repo root (no worktree) | Yes | No | | Bring your own PTY-based agent | Yes | No | | Per-workspace macOS sandbox | Yes | No | | Spotlight (live sync, detached HEAD) | Yes | Checkpoints onto branch | | Platforms | macOS + Linux (AppImage); Windows build from source | macOS only |
If you already pay for Claude Pro or Max, Termic uses the same quota as running claude in iTerm — no per-token markup.
Architecture
- Rust/Tauri 2 for the desktop runtime and IPC
- React for the UI
- xterm.js + WebGL for the terminal pane
- CodeMirror 6 for the diff view and editor
- sandbox-exec (macOS Seatbelt) for filesystem isolation
- In-process CONNECT proxy for network allowlisting
Sessions auto-resume on restart. Config is persisted in .termic.yaml files that can be committed to a repo.
FAQ
Q: Does this work with Claude Pro or Max only?
A: It works with any plan that includes the interactive claude CLI. The agent sees the same auth it would in iTerm. For CodeX, Copilot, etc., the respective CLI must be installed and authenticated on your system.
Q: Is there a Windows binary? A: Not yet. Build from source works today on Windows 11 with WebView2. The roadmap includes prebuilt MSI binaries.
Q: How does the sandbox work on Linux?
A: It does not — sandbox-exec is macOS-only. On Linux the Shield toggle is disabled. The roadmap includes bubblewrap/landlock support.
Q: Can I use local models via Ollama?
A: Yes. Termic’s agent registry accepts any CLI that runs in a terminal, including Ollama. The roadmap mentions first-class opencode and pi.dev support.
Conclusion
Termic solves a real pain point that emerged after June 2026 Anthropic SDK pricing changes: running Claude Code through a vendor SDK incurs a separate metered cost on top of your existing Pro/Max subscription. By spawning the CLI directly in a PTY, Termic stays on your existing plan with zero markup.
The git worktree isolation is the other key win — you can run four agents on the same branch in parallel without them stepping on each other, then Spotlight the best result back into your checkout. For teams that rely heavily on AI coding agents, that workflow difference alone is significant.
Install via Homebrew on macOS, or grab the AppImage on Linux. Source is on GitHub under AGPL-3.0.
Source: termic.dev | GitHub | HN: news.ycombinator.com/item?id=49056316
Related Posts
ai-setup
Recall – Persistent Memory for Claude Code via MCP Hooks
Recall gives Claude Code a permanent memory store that survives session restarts and context compaction. Four hooks capture and restore context automatically — with cloud SaaS or self-hosted options.
2/28/2026
dev-tools
Automotive Skills Suite for AI Engineering
Evaluate Automotive Skills Suite for APQP, ASPICE, HARA, safety-plan, and DIA workflows with setup notes, governance risks, and SME review guidance.
5/28/2026
dev-tools
awesome-agentic-ai-zh Roadmap Guide
Explore awesome-agentic-ai-zh as a Chinese agentic AI learning roadmap, with setup notes, track selection, study workflow, and evaluation guidance.
5/28/2026