NeuralCodeDOCS
Getting started

NeuralCode documentation

NeuralCode runs a team of Claude Code agents in parallel on your Mac — each in its own isolated git worktree. Start a task, let an agent work, review the diff, run it in the terminal, then merge. This documentation covers every feature in depth.

NeuralCode
Projects
2xlabs main
fix-test-bug
code-agents/fix-test-bug
enhance the hero text
code-agents/enhance-the-her…
add-dark-mode
code-agents/add-dark-mode
lyni main
No workspaces yet
dograh main
No workspaces yet
History4
Usage
Steve de silva
@steve
NeuralCode v0.1.0-beta.1BETA
Y
You
Add a section called ‘Cipher’ to the landing page. Leave it blank for now.
Claude Code
6 tool calls
Readapp/page.tsx
Editapp/page.tsx
Bashnpm run typecheck
Done. Added a blank section-shell with id="cipher"so it's anchor-linkable.
2m 22s
Ask to make changes, @mention files, run /commands
Opus 4.8 Thinking off
Self-review
Diff 3 Files
Mpage.tsxapp/+1 2
Acipher.cssstyles/+14
Mnav.tsxcomponents/+3 1
app/page.tsx+1 −2
@@ -247,8 +247,7 @@ export default function IndexPage() {
247 {/* Headline */}
248 <h1 className="hero-display mt-7">
249- <span>I</span> that works without the complexity.
249+ <span>I</span> that works.
250 </h1>
251 {/* Subtext */}
Terminal Console
steve@MacBook-Pro fix-test-bug % npm run dev
> vite
VITE v7.0.4 ready in 412 ms
➜ Local: http://localhost:1420/
steve@MacBook-Pro fix-test-bug % _
The NeuralCode workspace — sidebar of parallel workspaces, the agent chat, a live diff, and an embedded terminal.

What NeuralCode is#

NeuralCode is a native macOS desktop app (built with Tauri v2 and React) that orchestrates the Claude Code CLI. Instead of running one agent in your terminal, you run many at once — each agent works inside its own git worktree on its own branch, so they never step on each other. You supervise the results as diffs and merge what you like.

The parallel-agents model#

Everything in NeuralCode is built around three nested ideas:

ConceptWhat it is
ProjectA git repository you've opened. The sidebar groups work by project, and one project can hold many workspaces.
WorkspaceOne task. Backed by a dedicated git worktree on its own branch, with its own chat history, diff, terminal and review.
Agent runA single turn of the Claude Code agent inside a workspace — streamed live, with tool-use chips and a final summary.

Because each workspace is a real worktree, you can run a bug fix, a feature, and a refactor simultaneously and watch them all progress in the sidebar.

The end-to-end flow#

A typical loop looks like this:

  • Open a project — point NeuralCode at a git repo.
  • Create a workspace — name the task; NeuralCode spins up a worktree + branch.
  • Chat with the agent — describe the work; output streams in with tool-use chips.
  • Review — inspect the diff, let a second agent self-review, and run your tests.
  • Ship — commit, push, open a PR, or merge the branch into its base.
New here?
Start with Install & setup to get the app and its dependencies running, then read Workspaces & worktrees — the single most important concept in NeuralCode.

How these docs are organized#

  • Core workflow — the day-to-day: projects, workspaces, chat, diff, review, terminal, and shipping.
  • Integrations — GitHub (sign-in, issues, PRs) and Google sign-in.
  • Configuration — settings, models, custom commands, and permission modes.
  • Usage & maintenance — spend tracking, shortcuts, and updates.
  • Developer reference — the Tauri commands, streaming events, and data model under the hood.

Use the search box (⌘K) any time to jump straight to a feature.