Meta-Agent Building System

An autonomous AI system that builds, tests, and deploys production-ready AI agents from plain English requirements.

Overview

The Meta-Agent Factory takes natural language requirements and automatically:

• Generates complete agent code with LangGraph/CrewAI/Direct LLM
• Creates comprehensive test suites
• Builds Docker containers
• Generates Kubernetes deployments with auto-scaling
• Implements monitoring and observability
• Handles multi-tenancy and cost tracking

Key Features

• Autonomous Test-Fix Loop: Automatically fixes code until all tests pass
• Production-First: Every agent is deployment-ready from day one
• Scale-to-Zero: Agents cost $0 when idle using KEDA autoscaling
• Multi-Framework: Supports LangGraph, CrewAI, and Direct LLM patterns
• Observable: Built-in OpenTelemetry tracing and Prometheus metrics

Repository Structure

See REPO_LAYOUT.md for the complete directory structure.

Getting Started

# Clone the repository
git clone https://github.com/Instabidsai/Meta-agent-building-system.git

# Navigate to the project
cd Meta-agent-building-system

# Start building agents!

Example Usage

"Build me an agent that monitors Slack for questions about our product and answers them using our knowledge base"

The system will:

1. Identify required secrets (SLACK_TOKEN, OPENAI_API_KEY)
2. Create repository: instabids-agents/slack-support-bot
3. Generate complete agent code with worker loop
4. Create and run comprehensive tests
5. Build and validate Docker container
6. Generate Kubernetes manifests with autoscaling
7. Deliver production-ready system

Documentation

• Agent Build Runbook
• Model Selection Guide
• Session Logs

Built Agents

Check the examples/ directory for agents built by this system.