Juggernaut Labs Documentation
Juggernaut Labs is a code-first operating system for automating internal business processes using Large Language Models (LLMs). Unlike visual workflow builders, Juggernaut provides a declarative JSON API for defining complex decision trees, combined with React SDK interfaces and a TypeScript Plugin SDK for extensibility.
Core Philosophy
Processes as Code: Business logic is defined through structured JSON schemas (InputProcess API), version-controlled, and deployed programmatically. This enables:
- Reproducibility: Same input always produces same execution path
- Collaboration: Technical and non-technical stakeholders contribute via prompt engineering and schema design
- Integration: Native connection to databases, APIs, and external systems through type-safe plugins
Architecture Overview
┌─────────────────────────────────────────────────────────────┐
│ INPUTPROCESS API │
│ (JSON Workflow Definitions) │
└──────────────────────┬──────────────────────────────────────┘
│
┌─────────────▼──────────────┐
│ PIPELINES[] │ ← Sequential execution array
│ (Sync/Async/Conditional)│
└─────────────┬──────────────┘
│
┌────────▼────────┐
│ STEPS[] │ ← Individual AI tasks
│ (LLM/Plugin) │
└────────┬────────┘
│
┌─────────────▼──────────────┐
│ @juggernautlabs/views │ ← React SDK for interfaces
│ @juggernautlabs/plugins│ ← TypeScript plugin SDK
└────────────────────────────┘Key Components
1. InputProcess API
The declarative engine. You define workflows as JSON objects containing:
- Pipelines: Execution phases (array order = execution order)
- Steps: Individual tasks with LLM prompts or plugin calls
- State Management: Type-safe variable passing between steps
2. Views SDK (@juggernautlabs/views)
React toolkit for building Process-Driven Applications—interfaces that consume process outputs. Features:
- Automatic job data injection via hooks
- Type-safe output consumption
- Vibe-coding ready (standard React patterns)
3. Plugins SDK (@juggernautlabs/plugins)
TypeScript framework for extending capabilities:
- Database connectors (MongoDB, Postgres, etc.)
- External API integrations
- Custom business logic execution
- State modification during process runs
Workflow Lifecycle
- Define: Write InputProcess JSON specifying inputs, pipelines, steps, and output schemas
- Upload: Deploy process definition via API
- Execute: Trigger via API call or SDK with runtime inputs
- Interface: Build React view using
@juggernautlabs/viewsto display results - Extend: Develop plugins using
@juggernautlabs/pluginsfor custom integrations
When to Use Juggernaut
Ideal for:
- Multi-step AI workflows requiring structured output at each phase
- Processes needing conditional logic (if X > 0.8, route to specialist pipeline)
- Bulk operations (iterating over arrays with AI processing per item)
- Applications requiring type-safe database integration
- Teams wanting version-controlled business logic
Not a replacement for:
- Simple single-prompt chatbots (overkill)
- Visual drag-and-drop automation (Zapier/Make territory—though Juggernaut can orchestrate those via plugins)
Next Steps
- Core Concepts - Understand Pipelines, Steps, and State
- InputProcess API Reference - Complete JSON schema documentation
- Views SDK Setup - Build React interfaces for your processes
- Plugins SDK Guide - Develop custom integrations