Welcome to the Autonomous AI Agent project by @0xmooncity and the mooncitydev .
This project aims to create an autonomous AI-driven agent that can manage social media accounts (starting with Twitter), interact with users, remember context, handle tasks like sending tips (cryptocurrency microtransactions), and more—all with minimal human intervention. The long-term vision is to expand beyond Twitter to multiple social media platforms.
We're building this in TypeScript for type safety, extensive ecosystem support, and rapid development capabilities.
-
Autonomous Social Media Management
The agent should:- Read mentions, replies, and direct messages
- Generate and publish tweets autonomously using integrated LLMs (e.g., Claude API)
- Maintain a rolling memory and conversational context
- Monitor specific topics and hashtags (e.g., $BTB)
- Engage with community through likes and replies
-
Cryptocurrency Tipping and Transactions
The agent can:- Generate and securely store a wallet address and private keys
- Send small crypto tips to users who meet certain criteria
- Interact with blockchain APIs using ethers.js
- Track transaction history and user interactions
- Implement secure key management and encryption
-
Scalable and Extensible Architecture
Our code aims to:- Be easily extensible to other social media platforms
- Allow integration with various LLM providers and conversation strategies
- Encourage community contributions and improvements
- Support multiple blockchain networks and token standards
-
TypeScript Backend:
- Strong type system for enhanced code reliability
- Modern async/await patterns for efficient I/O operations
- Rich ecosystem of NPM packages
- Easy integration with Node.js
-
LLM Integration (Claude):
- Use language model APIs for natural language understanding and generation
- Context-aware responses with conversation history
- Custom prompt engineering for specific use cases
- Fallback mechanisms for API failures
-
Secure Storage of Keys and State:
- Environment-based configuration using dotenv
- Private keys and secrets are never hardcoded
- Secure key management practices
- Regular state persistence
-
Event-Driven Architecture:
- Periodically fetch mentions from Twitter
- Process each event through the LLM for intelligent responses
- Rate limiting and exponential backoff
- Error handling and retry mechanisms
-
Compliance and Safety Controls:
- Guardrails at the prompt and code level
- Content filtering and moderation
- Rate limiting and anti-spam measures
- Audit logging for all operations
-
Current Implementation:
Features implemented:- Twitter API integration using twitter-api-v2
- Claude API integration for LLM capabilities
- Ethereum wallet management with ethers.js
- State management and conversation tracking
- Rate limiting and error handling
-
Contributions Needed:
Priority areas:- Enhanced conversation strategies
- Additional social media platform integrations
- Improved security measures
- Testing and documentation
- UI/UX for monitoring and configuration
-
Node.js and npm:
- Node.js (v16 or later) from https://nodejs.org/
- Required packages listed in
package.json
-
Twitter Developer Account:
Required credentials:- API Key and Secret
- Access Token and Secret
- App permissions for read/write access
-
Claude API Key:
- Claude API key from Anthropic
- Rate limits and usage quotas consideration
src/
├── index.ts # Main application entry point
├── services/ # Core services
│ ├── TwitterApi.ts # Twitter API integration
│ ├── LLMClient.ts # Claude API integration
│ └── WalletManager.ts # Ethereum wallet management
└── types/ # TypeScript type definitions
└── index.ts # Shared types and interfaces
The application uses environment variables for configuration:
TWITTER_API_KEY- Twitter API keyTWITTER_API_SECRET- Twitter API secretTWITTER_ACCESS_TOKEN- Twitter access tokenTWITTER_ACCESS_TOKEN_SECRET- Twitter access token secretCLAUDE_API_KEY- Claude API keyWALLET_PRIVATE_KEY- Ethereum wallet private keyETH_RPC_URL- Ethereum RPC URL
We welcome contributions! Please see our Contributing Guidelines for details.
This project is licensed under the MIT License - see the LICENSE file for details.