Open source. Made for builders.

Run agents
in parallel.

Keep the final say.

Your AI coding agents, side by side. Each task gets its own git worktree. Review the changes. Choose what ships.

Free for macOS & Linux · Bring your own agents

A team of agents. One place to lead them.Ctrl+Tab in action / 0:22
Answer waiting agents. Review what’s ready. Silent demo.Explore the workflow

Your favorite agents. One place to run them.

  • Claude Code
  • Codex CLI
  • OpenCode
  • Copilot CLI
  • Antigravity CLI
Open source. MIT licensed. Yours to build with.

From prompt to reviewed code

More work in motion.
Less on your mind.

01Give each idea room

A task. A branch. A place to work.

Send one agent after a bug while another builds your next feature. Each task gets its own branch and worktree, keeping active working copies separate.

A focused task with its branch, changed files, notes and agent chat
From the workspace View full screenshot

02Keep the final say

Their changes. Your call.

Read the diff. Leave a comment. Send it back to the agent. When you're happy with the result, choose what to merge.

An inline review comment asking the agent to label forecast days by weekday
From the workspace View full screenshot

03See beyond the diff

Follow the change. Ask better questions.

Take a guided tour of a changed file or a whole branch. Ask follow-up questions at any stop, with the relevant code right beside you.

A guided change tour explaining a date-label issue beside the highlighted code
From the workspace View full screenshot

Fits the way you build.

  • Your editor stays yours

    Keep VS Code, Cursor, JetBrains, or Sublime.

  • Terminal or chat

    Run agent CLIs or use built-in Claude Code and Codex chat.

  • Ideas you can explore

    Mind maps and reasoning graphs, alongside your tasks.

  • Agents that delegate

    Subtasks in separate worktrees, with results you can review.

  • Desktop to phone

    Follow terminals and chats over Wi-Fi or Tailscale.

  • Open source. Yours.

    MIT licensed. Bring your own API keys.

Less overhead, more ownership

Keep your tools.
Lose the busywork.

Your terminal agents and your editor, with the coordination taken care of.

Managing parallel agents manually compared with Parallel Code
The workflowManaging it yourselfWith Parallel Code
Starting a taskCreate a branch and set up a working copy.Describe the task. Get a branch and worktree.
Following progressSwitch between individual terminal sessions.Keep your agent sessions in one workspace.
Reviewing the resultFind the right branch and inspect its changes.Open the task’s diff and leave inline comments.
Finishing upManage the branch and merge from your terminal.Review, then merge from the sidebar.
Explore tool comparisons

See it in action

Take a closer look.

Explore the workspace at your own pace: task views, inline reviews, guided tours, and room to think.

Run a team of AI coding agentsFull workflow tour 3:16
Meet Parallel Code Narrated introduction · 2:55
Meet Parallel CodeNarrated introduction 2:55
Agents working, waiting for input, and ready to merge
Agents working, waiting for input, and ready to mergeFull size
One task in focus with its chat, changed files, notes, and shell
One task in focus with its chat, changed files, notes, and shellFull size
Inline review comments you can send back to the agent
Inline review comments you can send back to the agentFull size
A guided tour of what the agent changed
A guided tour of what the agent changedFull size
The agent’s reasoning graph, updated live
The agent’s reasoning graph, updated liveFull size
A mind map with branches linked to live tasks
A mind map with branches linked to live tasksFull size
AI Arena — race agents head-to-head
AI Arena — race agents head-to-headFull size

A few more things

Good questions.
Clear answers.

Want to dig a little deeper?

Join the conversation on GitHub
Does it work with my IDE?

Parallel Code is a standalone Electron app. Keep using your preferred editor — VS Code, Cursor, JetBrains, Sublime, whatever you like. It manages the agents and worktrees; your IDE handles the code.

Which AI agents are supported?

Claude Code, Codex CLI, OpenCode, Copilot CLI, and Antigravity CLI out of the box. Any CLI-based coding agent that works in a terminal can be used.

How does the isolation work?

Each task gets its own git branch and worktree, so agents do not overwrite the same working copy. Branches can still conflict later if they change the same code; review and resolve those integration conflicts before merging back to main.

What platforms are supported?

macOS (universal .dmg) and Linux (.AppImage or .deb). Download from the GitHub releases page. There is no Windows build yet.

How is it different from Conductor or the Claude and Codex apps?

Each vendor's app runs that vendor's agent. Parallel Code puts Claude Code, Codex, Copilot and other terminal agents on one board, gives every task its own worktree, and runs the same review for all of them: inline comments sent back to the agent, a guided tour of the changes, and an optional review by a different model. Unlike Conductor, it is MIT licensed and runs on Linux as well as macOS. The comparison pages go into detail.

Can I use my Claude subscription?

For subscription use, sign in to the unmodified Claude Code CLI installed on your computer and use it in Parallel Code's terminal. The optional built-in Claude chat is a separate integration through Anthropic's Agent SDK. Anthropic's authentication guidance directs third-party SDK integrations to use a Claude Console API key or a supported cloud provider, and restricts third-party use of Claude subscription credentials. Check Anthropic's legal and compliance page for the current requirements.

Does my code go through a Parallel Code server?

No. Parallel Code has no server and no account. Your code goes only to the agent CLIs you run and to the models you choose in Settings. Phone access connects straight to your computer over Wi-Fi or Tailscale.

Who makes it, and what happens if it stops?

Johannes Millan, who has maintained the open-source task manager Super Productivity since 2016. Parallel Code is mostly a one-maintainer project with a few outside contributors, so that risk is real. It is limited by design: the code is MIT licensed and can be forked, the agent CLIs work without it, and every task is an ordinary git branch and worktree you can keep using.

Is it free?

Yes. Parallel Code is free and open source under the MIT license. Your API keys stay yours — nothing is proxied or collected.

Your next idea doesn't have to wait

Make room for
what’s next.

Put your agents to work. Keep the final say.

Free & open source · macOS & Linux · MIT licensed