Skip to main content

Welcome to Parallax

Parallax is a multi-agent orchestration framework for building reliable AI systems through consensus, voting, and confidence-scored aggregation.

Why Parallax?​

Single AI agents are unreliable. They hallucinate, make mistakes, and produce inconsistent results. Parallax solves this by orchestrating multiple agents to work together, using proven distributed systems patterns to achieve reliability.

Key Benefits​

  • Higher Accuracy - Multiple agents cross-validate each other's work
  • Confidence Scoring - Know how reliable each response is
  • Consensus Building - Aggregate multiple opinions into trusted results
  • Quality Gates - Filter out low-confidence responses automatically
  • Fault Tolerance - Graceful handling when individual agents fail

How It Works​

  1. Define a Pattern - Describe your orchestration flow in YAML or visually
  2. Register Agents - Connect your AI agents (any model, any provider)
  3. Execute - Parallax routes work to agents and aggregates results
  4. Get Results - Receive consensus results with confidence scores

Quick Example​

name: content-moderation
version: 1.0.0

input:
  content: string

agents:
  capabilities: [content-moderation]
  min: 3

execution:
  strategy: parallel

aggregation:
  strategy: voting
  method: majority

output:
  verdict: $vote.result
  confidence: $vote.confidence

What's Next?​

🚀 Get Started

Install Parallax and run your first pattern in under 5 minutes.

Quickstart Guide

📚 Learn Concepts

Understand agents, patterns, and how Parallax orchestrates them.

Core Concepts

Powered By​

Prism LogoPrism Logo
Prism DSL

Parallax patterns compile to Prism, a programming language where uncertainty is a first-class citizen. Learn more →

Open Source​

Parallax is open source under the Apache 2.0 license. The open source version includes:

  • ✅ Unlimited local agents
  • ✅ All orchestration patterns
  • ✅ YAML and visual pattern builder
  • ✅ Full SDK access
  • ✅ In-memory execution

Enterprise features add distributed execution, persistence, high availability, and more for production deployments.