Mushin Docs / Getting Started

Getting Started

Workbench is the development environment where you build applications with Mushin. It runs in your browser and is organized as a tabbed workspace — a bit like an IDE — with a project explorer, editors, and detail panels. This page gets you oriented.

Signing in

You reach Workbench at its address and choose Sign In. Workbench uses the same single sign-on that the rest of the Nandeshou platform uses, so you sign in through the identity service and are returned to the Workbench. After signing in you choose which organization (tenant) to work in — everything you build belongs to an organization and is isolated to it.

The shape of the Workbench

Once you are in, Workbench arranges itself around a central workspace of tabbed editors, framed by panels you can show and hide:

AreaWhat it holds
Explorer (left)The tree of your open project — its branches, design streams, and, once generated, its data model and pages
Workspace (center)The editors you open, as tabs — project settings, design streams, the BrainChain pipeline, and more
Details & PaletteContextual detail for whatever you have selected, and the building blocks you can add
Prompts (bottom)The prompts that guide the AI as it works

You open things from the Explorer or the left menu, and they appear as tabs you can arrange side by side.

Projects

Everything you build lives in a project. A project is one application. From the Projects area you create a new project, open an existing one, or edit its settings. Opening a project loads it into the Explorer and workspace so you can start working on it. See Projects for the details.

Two ways to build

When you create a project you choose how you want to work with it:

  • Conversational — you describe what you want in a chat and the assistant produces it. This is the quickest way to get an application running, and the best place to start. See Building Conversationally .
  • Neuro-symbolic (Mushin) — you build the application up as a versioned specification, with the AI’s natural-language design distilled into structured metadata you can review, branch, and commit. This is the deliberate, auditable way to build something larger. See The Neuro-Symbolic Approach .

Both produce the same kind of result — a generated Go backend and TypeScript frontend — through the same code generator .

The left menu

Beyond projects, the left menu gives you the platform’s other areas: Codegen Work and Codegen Runs for automated implementation (see Automating with the Autoloop ), Users and Directory for your organization, and your profile. These are covered in their own pages.