Shelly

A visual LLM flow builder. Create node-based prompt chains that execute sequentially against Claude and other LLM providers. All data stays client-side in IndexedDB.

Quickstart

# Install dependencies
npm install

# Start dev server
npm run dev

Open http://localhost:3000 in your browser.

Setup

You'll need an Anthropic API key. Enter it in the app settings — it's stored in localStorage (never sent to any server other than Anthropic's API).

Commands

npm run dev      # Dev server on localhost:3000
npm run build    # Production build
npm run start    # Serve production build
npm run lint     # ESLint check

Usage

1. Create a new flow from the sidebar
2. Add nodes to the canvas: user-input, agent, structured-output, router, output
3. Connect nodes by dragging edges between them
4. Configure each node (system prompts, templates, schemas, routing conditions)
5. Use {{variable}} syntax in templates to reference upstream node outputs
6. Run the flow — nodes execute sequentially from root to leaf

Node Types

Node	Purpose
user-input	Entry point — passes user input forward
agent	LLM call with system prompt + message template (optionally with VFS tool-call loop)
structured-output	LLM call returning JSON matching a defined schema
router	Conditional branching (equals, contains, gt, lt)
output	Terminal node for displaying final results

Tech Stack

• Next.js 14 (App Router)
• ReactFlow (canvas)
• Zustand + IndexedDB (state & persistence)
• shadcn/ui + Tailwind CSS (UI)
• Anthropic SDK (LLM provider)

Changelog

See docs/CHANGELOG.md for version history.