MicroGPT Agent SDK

Production-ready TypeScript SDK for building AI agents with advanced memory, multi-agent orchestration, and comprehensive testing capabilities.

Why MicroGPT Agent SDK?

Building production AI agents requires more than just API calls. You need:

• Persistent Memory: Agents that remember conversations across sessions
• Multi-Agent Coordination: Orchestrate teams of specialized agents
• Production Testing: Test agents with real human feedback before deployment
• Protocol Flexibility: Support for MCP, ACP, and other agent protocols
• Provider Agnostic: Works with OpenAI, Anthropic, Ollama, and more

MicroGPT Agent SDK provides battle-tested infrastructure so you can focus on building intelligent agent behaviors, not reinventing the wheel.

Key Features

🧠 Advanced Memory System

• Hybrid Memory: Combines Redis (short-term), Supabase (long-term), and vector databases (semantic)
• Automatic Context Management: Agents remember relevant information across conversations
• Distributed Memory: Share memory across multiple agent instances

🤝 Multi-Agent Orchestration

• Message Bus: Event-driven communication between agents
• Distributed Locking: Prevent race conditions in multi-agent systems
• Shared Context: Agents collaborate with shared state
• Workflow Coordination: Chain agents into complex workflows

🧪 Human-in-the-Loop Testing

• HITL Framework: Test agents with real human feedback before production
• Test Orchestration: Coordinate complex test scenarios
• Performance Benchmarking: Measure latency, cost, and quality
• Integration Testing: Test entire agent workflows

🔌 Protocol Support

• MCP Adapter: Model Context Protocol integration
• ACP Adapter: Agent Communication Protocol
• FIPA Adapter: FIPA agent standards
• Custom Protocols: Extend with your own protocols

🎯 Provider Flexibility

• OpenAI: GPT-4, GPT-3.5-turbo, function calling
• Anthropic: Claude 3 Opus, Sonnet, Haiku
• Ollama: Local model support
• Unified Interface: Switch providers without code changes

Quick Start

Installation

npm install microgpt-agent-sdk

Basic Agent

import { AgentSDK } from 'microgpt-agent-sdk';

const sdk = new AgentSDK({
  provider: 'openai',
  apiKey: process.env.OPENAI_API_KEY
});

const agent = sdk.createAgent({
  name: 'assistant',
  systemPrompt: 'You are a helpful assistant.'
});

const response = await agent.chat('Hello!');
console.log(response);

Agent with Memory

import { AgentSDK, MemoryConfig } from 'microgpt-agent-sdk';

const sdk = new AgentSDK({
  provider: 'openai',
  apiKey: process.env.OPENAI_API_KEY,
  memory: {
    shortTerm: {
      type: 'redis',
      url: process.env.REDIS_URL
    },
    longTerm: {
      type: 'supabase',
      url: process.env.SUPABASE_URL,
      key: process.env.SUPABASE_KEY
    }
  }
});

const agent = sdk.createAgent({
  name: 'assistant',
  systemPrompt: 'You are a helpful assistant with memory.',
  memoryEnabled: true
});

// Agent remembers context across sessions
await agent.chat('My name is Alice');
// Later...
await agent.chat('What is my name?'); // "Your name is Alice"

Multi-Agent System

import { Orchestrator } from 'microgpt-agent-sdk';

const orchestrator = new Orchestrator();

// Create specialized agents
const researcher = orchestrator.createAgent({
  name: 'researcher',
  systemPrompt: 'You research topics thoroughly.'
});

const writer = orchestrator.createAgent({
  name: 'writer',
  systemPrompt: 'You write clear, engaging content.'
});

// Coordinate workflow
const workflow = orchestrator.createWorkflow()
  .step('research', researcher, { task: 'Research AI safety' })
  .step('write', writer, {
    task: 'Write article',
    context: (results) => results.research
  });

const result = await workflow.execute();

Human-in-the-Loop Testing

import { TestFramework } from 'microgpt-agent-sdk';

const testFramework = new TestFramework();

// Create test suite
const suite = testFramework.createSuite({
  name: 'Customer Support Agent',
  agent: customerSupportAgent
});

// Add test scenarios
suite.addScenario({
  name: 'Handle refund request',
  input: 'I want a refund for order #12345',
  expectedBehavior: 'Agent should verify order and process refund',
  humanReviewer: true // Request human review
});

// Run tests
const results = await suite.run();
console.log(`Pass rate: ${results.passRate}%`);

Architecture

┌─────────────────────────────────────────────────────────────┐
│                     MicroGPT Agent SDK                       │
├─────────────────────────────────────────────────────────────┤
│                                                               │
│  ┌──────────────┐  ┌──────────────┐  ┌──────────────┐      │
│  │    Memory    │  │ Orchestration│  │   Testing    │      │
│  │              │  │              │  │              │      │
│  │ • Short-term │  │ • Message Bus│  │ • HITL       │      │
│  │ • Long-term  │  │ • Workflows  │  │ • Integration│      │
│  │ • Semantic   │  │ • Locking    │  │ • Performance│      │
│  └──────────────┘  └──────────────┘  └──────────────┘      │
│                                                               │
│  ┌──────────────┐  ┌──────────────┐  ┌──────────────┐      │
│  │  Protocols   │  │   Providers  │  │   Learning   │      │
│  │              │  │              │  │              │      │
│  │ • MCP        │  │ • OpenAI     │  │ • Patterns   │      │
│  │ • ACP        │  │ • Anthropic  │  │ • Optimization│     │
│  │ • FIPA       │  │ • Ollama     │  │ • Analytics  │      │
│  └──────────────┘  └──────────────┘  └──────────────┘      │
│                                                               │
└─────────────────────────────────────────────────────────────┘

Documentation

• Getting Started Guide
• Memory System
• Multi-Agent Orchestration
• Testing Framework
• Protocol Adapters
• API Reference

Examples

Check out the examples directory for complete working examples:

• Basic Agent
• Memory System
• Multi-Agent Orchestration
• HITL Testing
• Protocol Integration

Contributing

We welcome contributions! Please see CONTRIBUTING.md for guidelines.

License

MIT License - see LICENSE

💬 Community

Join our community to ask questions, share ideas, and connect with other developers building production-ready AI agents!

• GitHub Discussions - Ask questions, share your work, and discuss best practices
• GitHub Issues - Bug reports and feature requests
• Email: [email protected]

We're building a supportive community where developers help each other create robust, production-ready AI agents. Whether you're just getting started or scaling to production, your questions and contributions are welcome!

---

Built with ❤️ by Cognio Lab

MicroGPT Agent SDK is built and maintained by Cognio Lab, creators of production AI agent infrastructure.