🧠 OAS Multi-Engine Playground

Interactive demonstration of Open Agent Spec (OAS) with Cortex cost optimization - a command-line tool for generating AI agent projects from YAML specifications. Experience the complete workflow: define agent specs, generate code, test across multiple LLM engines, and optimize costs with intelligent routing and fallback strategies.

📝 YAML Specifications🛠️ Agent Generation🧠 Multi-Engine Support⚡ CLI Workflow💰 Cortex Cost Optimization

Open Agent Spec Playground

Generate AI agents from YAML specifications using the OAS CLI workflow

Agent Templates

Select an agent template to generate and test

Security Analyzer Agent

Security

AI agent specialized in security vulnerability analysis

oas init --spec security_analyzer.yaml --output security-agent/

Data Science Agent

Analytics

AI agent for data analysis and insights generation

oas init --spec data_scientist.yaml --output data-agent/

Content Creator Agent

Marketing

AI agent specialized in marketing content generation

oas init --spec content_creator.yaml --output content-agent/

Code Review Agent

Development

AI agent for automated code quality and best practices review

oas init --spec code_reviewer.yaml --output code-agent/

Research Assistant Agent

Research

AI agent for comprehensive research and fact-checking

oas init --spec research_assistant.yaml --output research-agent/

OAS Engine Configuration

Configure which LLM engines to generate agents for

OpenAI GPT-4

OpenAI

~3s

Claude 3.5 Sonnet

Anthropic

~3s

Grok 3

xAI

~2s

Local Model

Local

~5s

Custom Router

Custom

~3s

Cortex Optimizer

Cortex

~2s

Auto-generate Agents

Agent YAML Specification

Complete OAS agent specification for Security Analyzer Agent

name: "SecurityAnalyzer"
version: "1.0.0"
description: "AI agent for security vulnerability analysis"

intelligence:
  type: "llm"
  engine: "claude"
  model: "claude-3-5-sonnet-20241022"
  config:
    temperature: 0.3
    max_tokens: 1000

task:
  input:
    code_snippet: "string"
    language: "string"
  output:
    risk_level: "high|medium|low"
    vulnerabilities: "array"
    recommendations: "array"
    confidence: "number"

behavioral_contract:
  version: "0.1.2"
  description: "Security-focused contract with strict validation"
  behavioral_flags:
    conservatism: "high"
    temperature_control:
      mode: "strict"
      range: [0.1, 0.5]
  response_contract:
    output_format:
      required_fields: ["risk_level", "vulnerabilities", "recommendations", "confidence"]
    safety_checks:
      harmful_content: true
      pii_protection: true

Expected Features:

SQL injection detection

Security recommendations

Risk scoring