For devs running Claude Code, Codex, Gemini CLI, Aider, Cline — in parallel

The terminal IDE
for your AI agents.

Run AI agents in parallel — each in its own git worktree. Use your existing Claude or Codex subscription. Control the whole thing from your phone. Open source, native to macOS.

Download for macOS Star on GitHub
● Use your subscription ● Open source ● macOS · iOS ● No telemetry
Kickoff

All your agents in one tap.

Quick Actions are saved agent launches — name, prompt, model, even a starter command. One tap fires the whole set, each agent ready to work on its own task.

termloop · quick action · all agents in one tap
01 Open Quick Actions 02 One tap 03 Each on its own task
See it run

Four agents. Four tasks. One workspace.

Hand four agents four tasks at once. Each runs in its own worktree, boots its own dev server, opens its own PR. Your host checkout stays clean.

termloop · 4 agents · 1 loop · 18 seconds
01 Empty workspace 02 4 agents kick off 03 Apps boot with live URLs 04 PRs open 05 Pull, done
Ask-to

Stuck? Ask another agent.

When an agent hits a wall, it doesn't wait for you. The built-in ask_to MCP tool lets it consult any other running agent — different model, different worktree, different perspective. Cross-pollination, no human in the loop.

termloop · ask_to · cross-agent consult
01 Agent prepares the question 02 New agent picks it up 03 Back to the asker 04 Answer arrives
More inside

Built for serious agent workflows.

TermLoop sidebar showing worktree agents, open and merged PRs
Sidebar

Every thread, one rail.

Branch, PR status, dirty file count, port forwards, agent state — every workspace at a glance. Skim ten active threads without switching tabs.

Context Bank cataloging CLAUDE.md files across the project with usage scores
Context Bank

Shared, scored context.

TermLoop catalogs every CLAUDE.md, AGENTS.md, GEMINI.md across your tree, scores how much each one actually helps, and lets you clone the strong ones into other agents.

Questions

Frequently asked.

Do I pay separately for the AI?

No. TermLoop runs the agent CLIs you already use — Claude Code, Codex, Gemini CLI, Aider, Cline. You stay logged in with your existing subscription. TermLoop never proxies your API traffic and adds no billing layer.

Which agents and models can I run?

Anything with a CLI: Claude Code (Anthropic), Codex (OpenAI), Gemini CLI (Google), Aider, Cline, local Ollama, or your own scripts. TermLoop just hands each one a worktree and a pane — it's model-agnostic.

Is TermLoop a fork of something?

TermLoop builds on cmux, an open-source GPL-3.0 terminal multiplexer by Manaflow. cmux gives us the multiplexer foundation — splits, panes, the renderer. The agent-loop layer, worktree orchestration, ask_to MCP, Context Bank, mobile bridge, and per-project Claude accounts — all of it is what TermLoop adds on top.

Is it really open source?

Yes. The macOS app is a free download and the source is public. Inspect the socket protocol, the agent loop, the worktree orchestration — everything. No telemetry, no phone-home.

Linux or Windows?

macOS only today. Linux is on the roadmap. Windows isn't, for now — we wanted a native, low-latency UI before we wanted broad reach.

Does the iOS app actually work?

Yes — the iOS client connects to your TermLoop session over an authenticated TCP socket on your local network. It's stable today but still under active development, so expect ongoing iteration. Desktop is the primary surface for now.

What about my code and credentials?

They never leave your machine. Agents call their providers directly with your own subscription. The only network call TermLoop itself makes is checking for app updates, which you can disable.